(mellow music) >> Hello everyone, welcome to theCUBE's coverage of AWS Startup Showcase. This is the closing panel session on AI machine learning, the top startups generating generative AI on AWS. It's a great panel. This is going to be the experts talking about riding the wave in generative AI. We got Ankur Mehrotra, who's the director and general manager of AI and machine learning at AWS, and Clem Delangue, co-founder and CEO of Hugging Face, and Ori Goshen, who's the co-founder and CEO of AI21 Labs. Ori from Tel Aviv dialing in, and rest coming in here on theCUBE. Appreciate you coming on for this closing session for the Startup Showcase. >> Thanks for having us. >> Thank you for having us. >> Thank you. >> I'm super excited to have you all on. Hugging Face was recently in the news with the AWS relationship, so congratulations. Open source, open science, really driving the machine learning. And we got the AI21 Labs access to the LLMs, generating huge scale live applications, commercial applications, coming to the market, all powered by AWS. So everyone, congratulations on all your success, and thank you for headlining this panel. Let's get right into it. AWS is powering this wave here. We're seeing a lot of push here from applications. Ankur, set the table for us on the AI machine learning. It's not new, it's been goin' on for a while. Past three years have been significant advancements, but there's been a lot of work done in AI machine learning. Now it's released to the public. Everybody's super excited and now says, "Oh, the future's here!" It's kind of been going on for a while and baking. Now it's kind of coming out. What's your view here? Let's get it started. >> Yes, thank you. So, yeah, as you may be aware, Amazon has been in investing in machine learning research and development since quite some time now. And we've used machine learning to innovate and improve user experiences across different Amazon products, whether it's Alexa or Amazon.com. But we've also brought in our expertise to extend what we are doing in the space and add more generative AI technology to our AWS products and services, starting with CodeWhisperer, which is an AWS service that we announced a few months ago, which is, you can think of it as a coding companion as a service, which uses generative AI models underneath. And so this is a service that customers who have no machine learning expertise can just use. And we also are talking to customers, and we see a lot of excitement about generative AI, and customers who want to build these models themselves, who have the talent and the expertise and resources. For them, AWS has a number of different options and capabilities they can leverage, such as our custom silicon, such as Trainium and Inferentia, as well as distributed machine learning capabilities that we offer as part of SageMaker, which is an end-to-end machine learning development service. At the same time, many of our customers tell us that they're interested in not training and building these generative AI models from scratch, given they can be expensive and can require specialized talent and skills to build. And so for those customers, we are also making it super easy to bring in existing generative AI models into their machine learning development environment within SageMaker for them to use. So we recently announced our partnership with Hugging Face, where we are making it super easy for customers to bring in those models into their SageMaker development environment for fine tuning and deployment. And then we are also partnering with other proprietary model providers such as AI21 and others, where we making these generative AI models available within SageMaker for our customers to use. So our approach here is to really provide customers options and choices and help them accelerate their generative AI journey. >> Ankur, thank you for setting the table there. Clem and Ori, I want to get your take, because the riding the waves, the theme of this session, and to me being in California, I imagine the big surf, the big waves, the big talent out there. This is like alpha geeks, alpha coders, developers are really leaning into this. You're seeing massive uptake from the smartest people. Whether they're young or around, they're coming in with their kind of surfboards, (chuckles) if you will. These early adopters, they've been on this for a while; Now the waves are hitting. This is a big wave, everyone sees it. What are some of those early adopter devs doing? What are some of the use cases you're seeing right out of the gate? And what does this mean for the folks that are going to come in and get on this wave? Can you guys share your perspective on this? Because you're seeing the best talent now leaning into this. >> Yeah, absolutely. I mean, from Hugging Face vantage points, it's not even a a wave, it's a tidal wave, or maybe even the tide itself. Because actually what we are seeing is that AI and machine learning is not something that you add to your products. It's very much a new paradigm to do all technology. It's this idea that we had in the past 15, 20 years, one way to build software and to build technology, which was writing a million lines of code, very rule-based, and then you get your product. Now what we are seeing is that every single product, every single feature, every single company is starting to adopt AI to build the next generation of technology. And that works both to make the existing use cases better, if you think of search, if you think of social network, if you think of SaaS, but also it's creating completely new capabilities that weren't possible with the previous paradigm. Now AI can generate text, it can generate image, it can describe your image, it can do so many new things that weren't possible before. >> It's going to really make the developers really productive, right? I mean, you're seeing the developer uptake strong, right? >> Yes, we have over 15,000 companies using Hugging Face now, and it keeps accelerating. I really think that maybe in like three, five years, there's not going to be any company not using AI. It's going to be really kind of the default to build all technology. >> Ori, weigh in on this. APIs, the cloud. Now I'm a developer, I want to have live applications, I want the commercial applications on this. What's your take? Weigh in here. >> Yeah, first, I absolutely agree. I mean, we're in the midst of a technology shift here. I think not a lot of people realize how big this is going to be. Just the number of possibilities is endless, and I think hard to imagine. And I don't think it's just the use cases. I think we can think of it as two separate categories. We'll see companies and products enhancing their offerings with these new AI capabilities, but we'll also see new companies that are AI first, that kind of reimagine certain experiences. They build something that wasn't possible before. And that's why I think it's actually extremely exciting times. And maybe more philosophically, I think now these large language models and large transformer based models are helping us people to express our thoughts and kind of making the bridge from our thinking to a creative digital asset in a speed we've never imagined before. I can write something down and get a piece of text, or an image, or a code. So I'll start by saying it's hard to imagine all the possibilities right now, but it's certainly big. And if I had to bet, I would say it's probably at least as big as the mobile revolution we've seen in the last 20 years. >> Yeah, this is the biggest. I mean, it's been compared to the Enlightenment Age. I saw the Wall Street Journal had a recent story on this. We've been saying that this is probably going to be bigger than all inflection points combined in the tech industry, given what transformation is coming. I guess I want to ask you guys, on the early adopters, we've been hearing on these interviews and throughout the industry that there's already a set of big companies, a set of companies out there that have a lot of data and they're already there, they're kind of tinkering. Kind of reminds me of the old hyper scaler days where they were building their own scale, and they're eatin' glass, spittin' nails out, you know, they're hardcore. Then you got everybody else kind of saying board level, "Hey team, how do I leverage this?" How do you see those two things coming together? You got the fast followers coming in behind the early adopters. What's it like for the second wave coming in? What are those conversations for those developers like? >> I mean, I think for me, the important switch for companies is to change their mindset from being kind of like a traditional software company to being an AI or machine learning company. And that means investing, hiring machine learning engineers, machine learning scientists, infrastructure in members who are working on how to put these models in production, team members who are able to optimize models, specialized models, customized models for the company's specific use cases. So it's really changing this mindset of how you build technology and optimize your company building around that. Things are moving so fast that I think now it's kind of like too late for low hanging fruits or small, small adjustments. I think it's important to realize that if you want to be good at that, and if you really want to surf this wave, you need massive investments. If there are like some surfers listening with this analogy of the wave, right, when there are waves, it's not enough just to stand and make a little bit of adjustments. You need to position yourself aggressively, paddle like crazy, and that's how you get into the waves. So that's what companies, in my opinion, need to do right now. >> Ori, what's your take on the generative models out there? We hear a lot about foundation models. What's your experience running end-to-end applications for large foundation models? Any insights you can share with the app developers out there who are looking to get in? >> Yeah, I think first of all, it's start create an economy, where it probably doesn't make sense for every company to create their own foundation models. You can basically start by using an existing foundation model, either open source or a proprietary one, and start deploying it for your needs. And then comes the second round when you are starting the optimization process. You bootstrap, whether it's a demo, or a small feature, or introducing new capability within your product, and then start collecting data. That data, and particularly the human feedback data, helps you to constantly improve the model, so you create this data flywheel. And I think we're now entering an era where customers have a lot of different choice of how they want to start their generative AI endeavor. And it's a good thing that there's a variety of choices. And the really amazing thing here is that every industry, any company you speak with, it could be something very traditional like industrial or financial, medical, really any company. I think peoples now start to imagine what are the possibilities, and seriously think what's their strategy for adopting this generative AI technology. And I think in that sense, the foundation model actually enabled this to become scalable. So the barrier to entry became lower; Now the adoption could actually accelerate. >> There's a lot of integration aspects here in this new wave that's a little bit different. Before it was like very monolithic, hardcore, very brittle. A lot more integration, you see a lot more data coming together. I have to ask you guys, as developers come in and grow, I mean, when I went to college and you were a software engineer, I mean, I got a degree in computer science, and software engineering, that's all you did was code, (chuckles) you coded. Now, isn't it like everyone's a machine learning engineer at this point? Because that will be ultimately the science. So, (chuckles) you got open source, you got open software, you got the communities. Swami called you guys the GitHub of machine learning, Hugging Face is the GitHub of machine learning, mainly because that's where people are going to code. So this is essentially, machine learning is computer science. What's your reaction to that? >> Yes, my co-founder Julien at Hugging Face have been having this thing for quite a while now, for over three years, which was saying that actually software engineering as we know it today is a subset of machine learning, instead of the other way around. People would call us crazy a few years ago when we're seeing that. But now we are realizing that you can actually code with machine learning. So machine learning is generating code. And we are starting to see that every software engineer can leverage machine learning through open models, through APIs, through different technology stack. So yeah, it's not crazy anymore to think that maybe in a few years, there's going to be more people doing AI and machine learning. However you call it, right? Maybe you'll still call them software engineers, maybe you'll call them machine learning engineers. But there might be more of these people in a couple of years than there is software engineers today. >> I bring this up as more tongue in cheek as well, because Ankur, infrastructure's co is what made Cloud great, right? That's kind of the DevOps movement. But here the shift is so massive, there will be a game-changing philosophy around coding. Machine learning as code, you're starting to see CodeWhisperer, you guys have had coding companions for a while on AWS. So this is a paradigm shift. How is the cloud playing into this for you guys? Because to me, I've been riffing on some interviews where it's like, okay, you got the cloud going next level. This is an example of that, where there is a DevOps-like moment happening with machine learning, whether you call it coding or whatever. It's writing code on its own. Can you guys comment on what this means on top of the cloud? What comes out of the scale? What comes out of the benefit here? >> Absolutely, so- >> Well first- >> Oh, go ahead. >> Yeah, so I think as far as scale is concerned, I think customers are really relying on cloud to make sure that the applications that they build can scale along with the needs of their business. But there's another aspect to it, which is that until a few years ago, John, what we saw was that machine learning was a data scientist heavy activity. They were data scientists who were taking the data and training models. And then as machine learning found its way more and more into production and actual usage, we saw the MLOps become a thing, and MLOps engineers become more involved into the process. And then we now are seeing, as machine learning is being used to solve more business critical problems, we're seeing even legal and compliance teams get involved. We are seeing business stakeholders more engaged. So, more and more machine learning is becoming an activity that's not just performed by data scientists, but is performed by a team and a group of people with different skills. And for them, we as AWS are focused on providing the best tools and services for these different personas to be able to do their job and really complete that end-to-end machine learning story. So that's where, whether it's tools related to MLOps or even for folks who cannot code or don't know any machine learning. For example, we launched SageMaker Canvas as a tool last year, which is a UI-based tool which data analysts and business analysts can use to build machine learning models. So overall, the spectrum in terms of persona and who can get involved in the machine learning process is expanding, and the cloud is playing a big role in that process. >> Ori, Clem, can you guys weigh in too? 'Cause this is just another abstraction layer of scale. What's it mean for you guys as you look forward to your customers and the use cases that you're enabling? >> Yes, I think what's important is that the AI companies and providers and the cloud kind of work together. That's how you make a seamless experience and you actually reduce the barrier to entry for this technology. So that's what we've been super happy to do with AWS for the past few years. We actually announced not too long ago that we are doubling down on our partnership with AWS. We're excited to have many, many customers on our shared product, the Hugging Face deep learning container on SageMaker. And we are working really closely with the Inferentia team and the Trainium team to release some more exciting stuff in the coming weeks and coming months. So I think when you have an ecosystem and a system where the AWS and the AI providers, AI startups can work hand in hand, it's to the benefit of the customers and the companies, because it makes it orders of magnitude easier for them to adopt this new paradigm to build technology AI. >> Ori, this is a scale on reasoning too. The data's out there and making sense out of it, making it reason, getting comprehension, having it make decisions is next, isn't it? And you need scale for that. >> Yes. Just a comment about the infrastructure side. So I think really the purpose is to streamline and make these technologies much more accessible. And I think we'll see, I predict that we'll see in the next few years more and more tooling that make this technology much more simple to consume. And I think it plays a very important role. There's so many aspects, like the monitoring the models and their kind of outputs they produce, and kind of containing and running them in a production environment. There's so much there to build on, the infrastructure side will play a very significant role. >> All right, that's awesome stuff. I'd love to change gears a little bit and get a little philosophy here around AI and how it's going to transform, if you guys don't mind. There's been a lot of conversations around, on theCUBE here as well as in some industry areas, where it's like, okay, all the heavy lifting is automated away with machine learning and AI, the complexity, there's some efficiencies, it's horizontal and scalable across all industries. Ankur, good point there. Everyone's going to use it for something. And a lot of stuff gets brought to the table with large language models and other things. But the key ingredient will be proprietary data or human input, or some sort of AI whisperer kind of role, or prompt engineering, people are saying. So with that being said, some are saying it's automating intelligence. And that creativity will be unleashed from this. If the heavy lifting goes away and AI can fill the void, that shifts the value to the intellect or the input. And so that means data's got to come together, interact, fuse, and understand each other. This is kind of new. I mean, old school AI was, okay, got a big model, I provisioned it long time, very expensive. Now it's all free flowing. Can you guys comment on where you see this going with this freeform, data flowing everywhere, heavy lifting, and then specialization? >> Yeah, I think- >> Go ahead. >> Yeah, I think, so what we are seeing with these large language models or generative models is that they're really good at creating stuff. But I think it's also important to recognize their limitations. They're not as good at reasoning and logic. And I think now we're seeing great enthusiasm, I think, which is justified. And the next phase would be how to make these systems more reliable. How to inject more reasoning capabilities into these models, or augment with other mechanisms that actually perform more reasoning so we can achieve more reliable results. And we can count on these models to perform for critical tasks, whether it's medical tasks, legal tasks. We really want to kind of offload a lot of the intelligence to these systems. And then we'll have to get back, we'll have to make sure these are reliable, we'll have to make sure we get some sort of explainability that we can understand the process behind the generated results that we received. So I think this is kind of the next phase of systems that are based on these generated models. >> Clem, what's your view on this? Obviously you're at open community, open source has been around, it's been a great track record, proven model. I'm assuming creativity's going to come out of the woodwork, and if we can automate open source contribution, and relationships, and onboarding more developers, there's going to be unleashing of creativity. >> Yes, it's been so exciting on the open source front. We all know Bert, Bloom, GPT-J, T5, Stable Diffusion, that work up. The previous or the current generation of open source models that are on Hugging Face. It has been accelerating in the past few months. So I'm super excited about ControlNet right now that is really having a lot of impact, which is kind of like a way to control the generation of images. Super excited about Flan UL2, which is like a new model that has been recently released and is open source. So yeah, it's really fun to see the ecosystem coming together. Open source has been the basis for traditional software, with like open source programming languages, of course, but also all the great open source that we've gotten over the years. So we're happy to see that the same thing is happening for machine learning and AI, and hopefully can help a lot of companies reduce a little bit the barrier to entry. So yeah, it's going to be exciting to see how it evolves in the next few years in that respect. >> I think the developer productivity angle that's been talked about a lot in the industry will be accelerated significantly. I think security will be enhanced by this. I think in general, applications are going to transform at a radical rate, accelerated, incredible rate. So I think it's not a big wave, it's the water, right? I mean, (chuckles) it's the new thing. My final question for you guys, if you don't mind, I'd love to get each of you to answer the question I'm going to ask you, which is, a lot of conversations around data. Data infrastructure's obviously involved in this. And the common thread that I'm hearing is that every company that looks at this is asking themselves, if we don't rebuild our company, start thinking about rebuilding our business model around AI, we might be dinosaurs, we might be extinct. And it reminds me that scene in Moneyball when, at the end, it's like, if we're not building the model around your model, every company will be out of business. What's your advice to companies out there that are having those kind of moments where it's like, okay, this is real, this is next gen, this is happening. I better start thinking and putting into motion plans to refactor my business, 'cause it's happening, business transformation is happening on the cloud. This kind of puts an exclamation point on, with the AI, as a next step function. Big increase in value. So it's an opportunity for leaders. Ankur, we'll start with you. What's your advice for folks out there thinking about this? Do they put their toe in the water? Do they jump right into the deep end? What's your advice? >> Yeah, John, so we talk to a lot of customers, and customers are excited about what's happening in the space, but they often ask us like, "Hey, where do we start?" So we always advise our customers to do a lot of proof of concepts, understand where they can drive the biggest ROI. And then also leverage existing tools and services to move fast and scale, and try and not reinvent the wheel where it doesn't need to be. That's basically our advice to customers. >> Get it. Ori, what's your advice to folks who are scratching their head going, "I better jump in here. "How do I get started?" What's your advice? >> So I actually think that need to think about it really economically. Both on the opportunity side and the challenges. So there's a lot of opportunities for many companies to actually gain revenue upside by building these new generative features and capabilities. On the other hand, of course, this would probably affect the cogs, and incorporating these capabilities could probably affect the cogs. So I think we really need to think carefully about both of these sides, and also understand clearly if this is a project or an F word towards cost reduction, then the ROI is pretty clear, or revenue amplifier, where there's, again, a lot of different opportunities. So I think once you think about this in a structured way, I think, and map the different initiatives, then it's probably a good way to start and a good way to start thinking about these endeavors. >> Awesome. Clem, what's your take on this? What's your advice, folks out there? >> Yes, all of these are very good advice already. Something that you said before, John, that I disagreed a little bit, a lot of people are talking about the data mode and proprietary data. Actually, when you look at some of the organizations that have been building the best models, they don't have specialized or unique access to data. So I'm not sure that's so important today. I think what's important for companies, and it's been the same for the previous generation of technology, is their ability to build better technology faster than others. And in this new paradigm, that means being able to build machine learning faster than others, and better. So that's how, in my opinion, you should approach this. And kind of like how can you evolve your company, your teams, your products, so that you are able in the long run to build machine learning better and faster than your competitors. And if you manage to put yourself in that situation, then that's when you'll be able to differentiate yourself to really kind of be impactful and get results. That's really hard to do. It's something really different, because machine learning and AI is a different paradigm than traditional software. So this is going to be challenging, but I think if you manage to nail that, then the future is going to be very interesting for your company. >> That's a great point. Thanks for calling that out. I think this all reminds me of the cloud days early on. If you went to the cloud early, you took advantage of it when the pandemic hit. If you weren't native in the cloud, you got hamstrung by that, you were flatfooted. So just get in there. (laughs) Get in the cloud, get into AI, you're going to be good. Thanks for for calling that. Final parting comments, what's your most exciting thing going on right now for you guys? Ori, Clem, what's the most exciting thing on your plate right now that you'd like to share with folks? >> I mean, for me it's just the diversity of use cases and really creative ways of companies leveraging this technology. Every day I speak with about two, three customers, and I'm continuously being surprised by the creative ideas. And the future is really exciting of what can be achieved here. And also I'm amazed by the pace that things move in this industry. It's just, there's not at dull moment. So, definitely exciting times. >> Clem, what are you most excited about right now? >> For me, it's all the new open source models that have been released in the past few weeks, and that they'll keep being released in the next few weeks. I'm also super excited about more and more companies getting into this capability of chaining different models and different APIs. I think that's a very, very interesting development, because it creates new capabilities, new possibilities, new functionalities that weren't possible before. You can plug an API with an open source embedding model, with like a no-geo transcription model. So that's also very exciting. This capability of having more interoperable machine learning will also, I think, open a lot of interesting things in the future. >> Clem, congratulations on your success at Hugging Face. Please pass that on to your team. Ori, congratulations on your success, and continue to, just day one. I mean, it's just the beginning. It's not even scratching the service. Ankur, I'll give you the last word. What are you excited for at AWS? More cloud goodness coming here with AI. Give you the final word. >> Yeah, so as both Clem and Ori said, I think the research in the space is moving really, really fast, so we are excited about that. But we are also excited to see the speed at which enterprises and other AWS customers are applying machine learning to solve real business problems, and the kind of results they're seeing. So when they come back to us and tell us the kind of improvement in their business metrics and overall customer experience that they're driving and they're seeing real business results, that's what keeps us going and inspires us to continue inventing on their behalf. >> Gentlemen, thank you so much for this awesome high impact panel. Ankur, Clem, Ori, congratulations on all your success. We'll see you around. Thanks for coming on. Generative AI, riding the wave, it's a tidal wave, it's the water, it's all happening. All great stuff. This is season three, episode one of AWS Startup Showcase closing panel. This is the AI ML episode, the top startups building generative AI on AWS. I'm John Furrier, your host. Thanks for watching. (mellow music)

(chill electronic music) >> Hello everyone, welcome to theCUBE's presentation of the AWS Startups Showcase, AI and machine learning, the top startups building generative AI on AWS. This is the season three, episode one of the ongoing series covering the exciting startups from the AWS ecosystem, talk about AI and machine learning. Can't believe it's three years and season one. I'm your host, John Furrier. Got a great guest today, we're joined by Joseph Nelson, the co-founder and CEO of Roboflow, doing some cutting edge stuff around computer vision and really at the front end of this massive wave coming around, large language models, computer vision. The next gen AI is here, and it's just getting started. We haven't even scratched a service. Thanks for joining us today. >> Thanks for having me. >> So you got to love the large language model, foundation models, really educating the mainstream world. ChatGPT has got everyone in the frenzy. This is educating the world around this next gen AI capabilities, enterprise, image and video data, all a big part of it. I mean the edge of the network, Mobile World Conference is happening right now, this month, and it's just ending up, it's just continue to explode. Video is huge. So take us through the company, do a quick explanation of what you guys are doing, when you were founded. Talk about what the company's mission is, and what's your North Star, why do you exist? >> Yeah, Roboflow exists to really kind of make the world programmable. I like to say make the world be read and write access. And our North Star is enabling developers, predominantly, to build that future. If you look around, anything that you see will have software related to it, and can kind of be turned into software. The limiting reactant though, is how to enable computers and machines to understand things as well as people can. And in a lot of ways, computer vision is that missing element that enables anything that you see to become software. So in the virtue of, if software is eating the world, computer vision kind of makes the aperture infinitely wide. It's something that I kind of like, the way I like to frame it. And the capabilities are there, the open source models are there, the amount of data is there, the computer capabilities are only improving annually, but there's a pretty big dearth of tooling, and an early but promising sign of the explosion of use cases, models, and data sets that companies, developers, hobbyists alike will need to bring these capabilities to bear. So Roboflow is in the game of building the community around that capability, building the use cases that allow developers and enterprises to use computer vision, and providing the tooling for companies and developers to be able to add computer vision, create better data sets, and deploy to production, quickly, easily, safely, invaluably. >> You know, Joseph, the word in production is actually real now. You're seeing a lot more people doing in production activities. That's a real hot one and usually it's slower, but it's gone faster, and I think that's going to be more the same. And I think the parallel between what we're seeing on the large language models coming into computer vision, and as you mentioned, video's data, right? I mean we're doing video right now, we're transcribing it into a transcript, linking up to your linguistics, times and the timestamp, I mean everything's data and that really kind of feeds. So this connection between what we're seeing, the large language and computer vision are coming together kind of cousins, brothers. I mean, how would you compare, how would you explain to someone, because everyone's like on this wave of watching people bang out their homework assignments, and you know, write some hacks on code with some of the open AI technologies, there is a corollary directly related to to the vision side. Can you explain? >> Yeah, the rise of large language models are showing what's possible, especially with text, and I think increasingly will get multimodal as the images and video become ingested. Though there's kind of this still core missing element of basically like understanding. So the rise of large language models kind of create this new area of generative AI, and generative AI in the context of computer vision is a lot of, you know, creating video and image assets and content. There's also this whole surface area to understanding what's already created. Basically digitizing physical, real world things. I mean the Metaverse can't be built if we don't know how to mirror or create or identify the objects that we want to interact with in our everyday lives. And where computer vision comes to play in, especially what we've seen at Roboflow is, you know, a little over a hundred thousand developers now have built with our tools. That's to the tune of a hundred million labeled open source images, over 10,000 pre-trained models. And they've kind of showcased to us all of the ways that computer vision is impacting and bringing the world to life. And these are things that, you know, even before large language models and generative AI, you had pretty impressive capabilities, and when you add the two together, it actually unlocks these kind of new capabilities. So for example, you know, one of our users actually powers the broadcast feeds at Wimbledon. So here we're talking about video, we're streaming, we're doing things live, we've got folks that are cropping and making sure we look good, and audio/visual all plugged in correctly. When you broadcast Wimbledon, you'll notice that the camera controllers need to do things like track the ball, which is moving at extremely high speeds and zoom crop, pan tilt, as well as determine if the ball bounced in or out. The very controversial but critical key to a lot of tennis matches. And a lot of that has been historically done with the trained, but fallible human eye and computer vision is, you know, well suited for this task to say, how do we track, pan, tilt, zoom, and see, track the tennis ball in real time, run at 30 plus frames per second, and do it all on the edge. And those are capabilities that, you know, were kind of like science fiction, maybe even a decade ago, and certainly five years ago. Now the interesting thing, is that with the advent of of generative AI, you can start to do things like create your own training data sets, or kind of create logic around once you have this visual input. And teams at Tesla have actually been speaking about, of course the autopilot team's focused on doing vision tasks, but they've combined large language models to add reasoning and logic. So given that you see, let's say the tennis ball, what do you want to do? And being able to combine the capabilities of what LLM's represent, which is really a lot of basically, core human reasoning and logic, with computer vision for the inputs of what's possible, creates these new capabilities, let alone multimodality, which I'm sure we'll talk more about. >> Yeah, and it's really, I mean it's almost intoxicating. It's amazing that this is so capable because the cloud scales here, you got the edge developing, you can decouple compute power, and let Moore's law and all the new silicone and the processors and the GPUs do their thing, and you got open source booming. You're kind of getting at this next segment I wanted to get into, which is the, how people should be thinking about these advances of the computer vision. So this is now a next wave, it's here. I mean I'd love to have that for baseball because I'm always like, "Oh, it should have been a strike." I'm sure that's going to be coming soon, but what is the computer vision capable of doing today? I guess that's my first question. You hit some of it, unpack that a little bit. What does general AI mean in computer vision? What's the new thing? Because there are old technology's been around, proprietary, bolted onto hardware, but hardware advances at a different pace, but now you got new capabilities, generative AI for vision, what does that mean? >> Yeah, so computer vision, you know, at its core is basically enabling machines, computers, to understand, process, and act on visual data as effective or more effective than people can. Traditionally this has been, you know, task types like classification, which you know, identifying if a given image belongs in a certain category of goods on maybe a retail site, is the shoes or is it clothing? Or object detection, which is, you know, creating bounding boxes, which allows you to do things like count how many things are present, or maybe measure the speed of something, or trigger an alert when something becomes visible in frame that wasn't previously visible in frame, or instant segmentation where you're creating pixel wise segmentations for both instance and semantic segmentation, where you often see these kind of beautiful visuals of the polygon surrounding objects that you see. Then you have key point detection, which is where you see, you know, athletes, and each of their joints are kind of outlined is another more traditional type problem in signal processing and computer vision. With generative AI, you kind of get a whole new class of problem types that are opened up. So in a lot of ways I think about generative AI in computer vision as some of the, you know, problems that you aimed to tackle, might still be better suited for one of the previous task types we were discussing. Some of those problem types may be better suited for using a generative technique, and some are problem types that just previously wouldn't have been possible absent generative AI. And so if you make that kind of Venn diagram in your head, you can think about, okay, you know, visual question answering is a task type where if I give you an image and I say, you know, "How many people are in this image?" We could either build an object detection model that might count all those people, or maybe a visual question answering system would sufficiently answer this type of problem. Let alone generative AI being able to create new training data for old systems. And that's something that we've seen be an increasingly prominent use case for our users, as much as things that we advise our customers and the community writ large to take advantage of. So ultimately those are kind of the traditional task types. I can give you some insight, maybe, into how I think about what's possible today, or five years or ten years as you sort go back. >> Yes, definitely. Let's get into that vision. >> So I kind of think about the types of use cases in terms of what's possible. If you just imagine a very simple bell curve, your normal distribution, for the longest time, the types of things that are in the center of that bell curve are identifying objects that are very common or common objects in context. Microsoft published the COCO Dataset in 2014 of common objects and contexts, of hundreds of thousands of images of chairs, forks, food, person, these sorts of things. And you know, the challenge of the day had always been, how do you identify just those 80 objects? So if we think about the bell curve, that'd be maybe the like dead center of the curve, where there's a lot of those objects present, and it's a very common thing that needs to be identified. But it's a very, very, very small sliver of the distribution. Now if you go out to the way long tail, let's go like deep into the tail of this imagined visual normal distribution, you're going to have a problem like one of our customers, Rivian, in tandem with AWS, is tackling, to do visual quality assurance and manufacturing in production processes. Now only Rivian knows what a Rivian is supposed to look like. Only they know the imagery of what their goods that are going to be produced are. And then between those long tails of proprietary data of highly specific things that need to be understood, in the center of the curve, you have a whole kind of messy middle, type of problems I like to say. The way I think about computer vision advancing, is it's basically you have larger and larger and more capable models that eat from the center out, right? So if you have a model that, you know, understands the 80 classes in COCO, well, pretty soon you have advances like Clip, which was trained on 400 million image text pairs, and has a greater understanding of a wider array of objects than just 80 classes in context. And over time you'll get more and more of these larger models that kind of eat outwards from that center of the distribution. And so the question becomes for companies, when can you rely on maybe a model that just already exists? How do you use your data to get what may be capable off the shelf, so to speak, into something that is usable for you? Or, if you're in those long tails and you have proprietary data, how do you take advantage of the greatest asset you have, which is observed visual information that you want to put to work for your customers, and you're kind of living in the long tails, and you need to adapt state of the art for your capabilities. So my mental model for like how computer vision advances is you have that bell curve, and you have increasingly powerful models that eat outward. And multimodality has a role to play in that, larger models have a role to play in that, more compute, more data generally has a role to play in that. But it will be a messy and I think long condition. >> Well, the thing I want to get, first of all, it's great, great mental model, I appreciate that, 'cause I think that makes a lot of sense. The question is, it seems now more than ever, with the scale and compute that's available, that not only can you eat out to the middle in your example, but there's other models you can integrate with. In the past there was siloed, static, almost bespoke. Now you're looking at larger models eating into the bell curve, as you said, but also integrating in with other stuff. So this seems to be part of that interaction. How does, first of all, is that really happening? Is that true? And then two, what does that mean for companies who want to take advantage of this? Because the old model was operational, you know? I have my cameras, they're watching stuff, whatever, and like now you're in this more of a, distributed computing, computer science mindset, not, you know, put the camera on the wall kind of- I'm oversimplifying, but you know what I'm saying. What's your take on that? >> Well, to the first point of, how are these advances happening? What I was kind of describing was, you know, almost uni-dimensional in that you have like, you're only thinking about vision, but the rise of generative techniques and multi-modality, like Clip is a multi-modal model, it has 400 million image text pairs. That will advance the generalizability at a faster rate than just treating everything as only vision. And that's kind of where LLMs and vision will intersect in a really nice and powerful way. Now in terms of like companies, how should they be thinking about taking advantage of these trends? The biggest thing that, and I think it's different, obviously, on the size of business, if you're an enterprise versus a startup. The biggest thing that I think if you're an enterprise, and you have an established scaled business model that is working for your customers, the question becomes, how do you take advantage of that established data moat, potentially, resource moats, and certainly, of course, establish a way of providing value to an end user. So for example, one of our customers, Walmart, has the advantage of one of the largest inventory and stock of any company in the world. And they also of course have substantial visual data, both from like their online catalogs, or understanding what's in stock or out of stock, or understanding, you know, the quality of things that they're going from the start of their supply chain to making it inside stores, for delivery of fulfillments. All these are are visual challenges. Now they already have a substantial trove of useful imagery to understand and teach and train large models to understand each of the individual SKUs and products that are in their stores. And so if I'm a Walmart, what I'm thinking is, how do I make sure that my petabytes of visual information is utilized in a way where I capture the proprietary benefit of the models that I can train to do tasks like, what item was this? Or maybe I'm going to create AmazonGo-like technology, or maybe I'm going to build like delivery robots, or I want to automatically know what's in and out of stock from visual input fees that I have across my in-store traffic. And that becomes the question and flavor of the day for enterprises. I've got this large amount of data, I've got an established way that I can provide more value to my own customers. How do I ensure I take advantage of the data advantage I'm already sitting on? If you're a startup, I think it's a pretty different question, and I'm happy to talk about. >> Yeah, what's startup angle on this? Because you know, they're going to want to take advantage. It's like cloud startups, cloud native startups, they were born in the cloud, they never had an IT department. So if you're a startup, is there a similar role here? And if I'm a computer vision startup, what's that mean? So can you share your your take on that, because there'll be a lot of people starting up from this. >> So the startup on the opposite advantage and disadvantage, right? Like a startup doesn't have an proven way of delivering repeatable value in the same way that a scaled enterprise does. But it does have the nimbleness to identify and take advantage of techniques that you can start from a blank slate. And I think the thing that startups need to be wary of in the generative AI enlarged language model, in multimodal world, is building what I like to call, kind of like sandcastles. A sandcastle is maybe a business model or a capability that's built on top of an assumption that is going to be pretty quickly wiped away by improving underlying model technology. So almost like if you imagine like the ocean, the waves are coming in, and they're going to wipe away your progress. You don't want to be in the position of building sandcastle business where, you don't want to bet on the fact that models aren't going to get good enough to solve the task type that you might be solving. In other words, don't take a screenshot of what's capable today. Assume that what's capable today is only going to continue to become possible. And so for a startup, what you can do, that like enterprises are quite comparatively less good at, is embedding these capabilities deeply within your products and delivering maybe a vertical based experience, where AI kind of exists in the background. >> Yeah. >> And we might not think of companies as, you know, even AI companies, it's just so embedded in the experience they provide, but that's like the vertical application example of taking AI and making it be immediately usable. Or, of course there's tons of picks and shovels businesses to be built like Roboflow, where you're enabling these enterprises to take advantage of something that they have, whether that's their data sets, their computes, or their intellect. >> Okay, so if I hear that right, by the way, I love, that's horizontally scalable, that's the large language models, go up and build them the apps, hence your developer focus. I'm sure that's probably the reason that the tsunami of developer's action. So you're saying picks and shovels tools, don't try to replicate the platform of what could be the platform. Oh, go to a VC, I'm going to build a platform. No, no, no, no, those are going to get wiped away by the large language models. Is there one large language model that will rule the world, or do you see many coming? >> Yeah, so to be clear, I think there will be useful platforms. I just think a lot of people think that they're building, let's say, you know, if we put this in the cloud context, you're building a specific type of EC2 instance. Well, it turns out that Amazon can offer that type of EC2 instance, and immediately distribute it to all of their customers. So you don't want to be in the position of just providing something that actually ends up looking like a feature, which in the context of AI, might be like a small incremental improvement on the model. If that's all you're doing, you're a sandcastle business. Now there's a lot of platform businesses that need to be built that enable businesses to get to value and do things like, how do I monitor my models? How do I create better models with my given data sets? How do I ensure that my models are doing what I want them to do? How do I find the right models to use? There's all these sorts of platform wide problems that certainly exist for businesses. I just think a lot of startups that I'm seeing right now are making the mistake of assuming the advances we're seeing are not going to accelerate or even get better. >> So if I'm a customer, if I'm a company, say I'm a startup or an enterprise, either one, same question. And I want to stand up, and I have developers working on stuff, I want to start standing up an environment to start doing stuff. Is that a service provider? Is that a managed service? Is that you guys? So how do you guys fit into your customers leaning in? Is it just for developers? Are you targeting with a specific like managed service? What's the product consumption? How do you talk to customers when they come to you? >> The thing that we do is enable, we give developers superpowers to build automated inventory tracking, self-checkout systems, identify if this image is malignant cancer or benign cancer, ensure that these products that I've produced are correct. Make sure that that the defect that might exist on this electric vehicle makes its way back for review. All these sorts of problems are immediately able to be solved and tackled. In terms of the managed services element, we have solutions as integrators that will often build on top of our tools, or we'll have companies that look to us for guidance, but ultimately the company is in control of developing and building and creating these capabilities in house. I really think the distinction is maybe less around managed service and tool, and more around ownership in the era of AI. So for example, if I'm using a managed service, in that managed service, part of their benefit is that they are learning across their customer sets, then it's a very different relationship than using a managed service where I'm developing some amount of proprietary advantages for my data sets. And I think that's a really important thing that companies are becoming attuned to, just the value of the data that they have. And so that's what we do. We tell companies that you have this proprietary, immense treasure trove of data, use that to your advantage, and think about us more like a set of tools that enable you to get value from that capability. You know, the HashiCorp's and GitLab's of the world have proven like what these businesses look like at scale. >> And you're targeting developers. When you go into a company, do you target developers with freemium, is there a paid service? Talk about the business model real quick. >> Sure, yeah. The tools are free to use and get started. When someone signs up for Roboflow, they may elect to make their work open source, in which case we're able to provide even more generous usage limits to basically move the computer vision community forward. If you elect to make your data private, you can use our hosted data set managing, data set training, model deployment, annotation tooling up to some limits. And then usually when someone validates that what they're doing gets them value, they purchase a subscription license to be able to scale up those capabilities. So like most developer centric products, it's free to get started, free to prove, free to poke around, develop what you think is possible. And then once you're getting to value, then we're able to capture the commercial upside in the value that's being provided. >> Love the business model. It's right in line with where the market is. There's kind of no standards bodies these days. The developers are the ones who are deciding kind of what the standards are by their adoption. I think making that easy for developers to get value as the model open sources continuing to grow, you can see more of that. Great perspective Joseph, thanks for sharing that. Put a plug in for the company. What are you guys doing right now? Where are you in your growth? What are you looking for? How should people engage? Give the quick commercial for the company. >> So as I mentioned, Roboflow is I think one of the largest, if not the largest collections of computer vision models and data sets that are open source, available on the web today, and have a private set of tools that over half the Fortune 100 now rely on those tools. So we're at the stage now where we know people want what we're working on, and we're continuing to drive that type of adoption. So companies that are looking to make better models, improve their data sets, train and deploy, often will get a lot of value from our tools, and certainly reach out to talk. I'm sure there's a lot of talented engineers that are tuning in too, we're aggressively hiring. So if you are interested in being a part of making the world programmable, and being at the ground floor of the company that's creating these capabilities to be writ large, we'd love to hear from you. >> Amazing, Joseph, thanks so much for coming on and being part of the AWS Startup Showcase. Man, if I was in my twenties, I'd be knocking on your door, because it's the hottest trend right now, it's super exciting. Generative AI is just the beginning of massive sea change. Congratulations on all your success, and we'll be following you guys. Thanks for spending the time, really appreciate it. >> Thanks for having me. >> Okay, this is season three, episode one of the ongoing series covering the exciting startups from the AWS ecosystem, talking about the hottest things in tech. I'm John Furrier, your host. Thanks for watching. (chill electronic music)

(upbeat music) >> Welcome everyone to theCUBE's presentation of the AWS Startup Showcase AI Machine Learning Top Startups Building Generative AI on AWS. This is season 3, episode 1 of the ongoing series covering the exciting startup from the AWS ecosystem to talk about AI and machine learning. I'm your host, John Furrier. I'm joined by two great guests here, Adam Wenchel, who's the CEO of Arthur, and Chief Scientist of Arthur, John Dickerson. Talk about how they help people build better LLM AI systems to get them into the market faster. Gentlemen, thank you for coming on. >> Yeah, thanks for having us, John. >> Well, I got to say I got to temper my enthusiasm because the last few months explosion of interest in LLMs with ChatGPT, has opened the eyes to everybody around the reality of that this is going next gen, this is it, this is the moment, this is the the point we're going to look back and say, this is the time where AI really hit the scene for real applications. So, a lot of Large Language Models, also known as LLMs, foundational models, and generative AI is all booming. This is where all the alpha developers are going. This is where everyone's focusing their business model transformations on. This is where developers are seeing action. So it's all happening, the wave is here. So I got to ask you guys, what are you guys seeing right now? You're in the middle of it, it's hitting you guys right on. You're in the front end of this massive wave. >> Yeah, John, I don't think you have to temper your enthusiasm at all. I mean, what we're seeing every single day is, everything from existing enterprise customers coming in with new ways that they're rethinking, like business things that they've been doing for many years that they can now do an entirely different way, as well as all manner of new companies popping up, applying LLMs to everything from generating code and SQL statements to generating health transcripts and just legal briefs. Everything you can imagine. And when you actually sit down and look at these systems and the demos we get of them, the hype is definitely justified. It's pretty amazing what they're going to do. And even just internally, we built, about a month ago in January, we built an Arthur chatbot so customers could ask questions, technical questions from our, rather than read our product documentation, they could just ask this LLM a particular question and get an answer. And at the time it was like state of the art, but then just last week we decided to rebuild it because the tooling has changed so much that we, last week, we've completely rebuilt it. It's now way better, built on an entirely different stack. And the tooling has undergone a full generation worth of change in six weeks, which is crazy. So it just tells you how much energy is going into this and how fast it's evolving right now. >> John, weigh in as a chief scientist. I mean, you must be blown away. Talk about kid in the candy store. I mean, you must be looking like this saying, I mean, she must be super busy to begin with, but the change, the acceleration, can you scope the kind of change you're seeing and be specific around the areas you're seeing movement and highly accelerated change? >> Yeah, definitely. And it is very, very exciting actually, thinking back to when ChatGPT was announced, that was a night our company was throwing an event at NeurIPS, which is maybe the biggest machine learning conference out there. And the hype when that happened was palatable and it was just shocking to see how well that performed. And then obviously over the last few months since then, as LLMs have continued to enter the market, we've seen use cases for them, like Adam mentioned all over the place. And so, some things I'm excited about in this space are the use of LLMs and more generally, foundation models to redesign traditional operations, research style problems, logistics problems, like auctions, decisioning problems. So moving beyond the already amazing news cases, like creating marketing content into more core integration and a lot of the bread and butter companies and tasks that drive the American ecosystem. And I think we're just starting to see some of that. And in the next 12 months, I think we're going to see a lot more. If I had to make other predictions, I think we're going to continue seeing a lot of work being done on managing like inference time costs via shrinking models or distillation. And I don't know how to make this prediction, but at some point we're going to be seeing lots of these very large scale models operating on the edge as well. So the time scales are extremely compressed, like Adam mentioned, 12 months from now, hard to say. >> We were talking on theCUBE prior to this session here. We had theCUBE conversation here and then the Wall Street Journal just picked up on the same theme, which is the printing press moment created the enlightenment stage of the history. Here we're in the whole nother automating intellect efficiency, doing heavy lifting, the creative class coming back, a whole nother level of reality around the corner that's being hyped up. The question is, is this justified? Is there really a breakthrough here or is this just another result of continued progress with AI? Can you guys weigh in, because there's two schools of thought. There's the, "Oh my God, we're entering a new enlightenment tech phase, of the equivalent of the printing press in all areas. Then there's, Ah, it's just AI (indistinct) inch by inch. What's your guys' opinion? >> Yeah, I think on the one hand when you're down in the weeds of building AI systems all day, every day, like we are, it's easy to look at this as an incremental progress. Like we have customers who've been building on foundation models since we started the company four years ago, particular in computer vision for classification tasks, starting with pre-trained models, things like that. So that part of it doesn't feel real new, but what does feel new is just when you apply these things to language with all the breakthroughs and computational efficiency, algorithmic improvements, things like that, when you actually sit down and interact with ChatGPT or one of the other systems that's out there that's building on top of LLMs, it really is breathtaking, like, the level of understanding that they have and how quickly you can accelerate your development efforts and get an actual working system in place that solves a really important real world problem and makes people way faster, way more efficient. So I do think there's definitely something there. It's more than just incremental improvement. This feels like a real trajectory inflection point for the adoption of AI. >> John, what's your take on this? As people come into the field, I'm seeing a lot of people move from, hey, I've been coding in Python, I've been doing some development, I've been a software engineer, I'm a computer science student. I'm coding in C++ old school, OG systems person. Where do they come in? Where's the focus, where's the action? Where are the breakthroughs? Where are people jumping in and rolling up their sleeves and getting dirty with this stuff? >> Yeah, all over the place. And it's funny you mentioned students in a different life. I wore a university professor hat and so I'm very, very familiar with the teaching aspects of this. And I will say toward Adam's point, this really is a leap forward in that techniques like in a co-pilot for example, everybody's using them right now and they really do accelerate the way that we develop. When I think about the areas where people are really, really focusing right now, tooling is certainly one of them. Like you and I were chatting about LangChain right before this interview started, two or three people can sit down and create an amazing set of pipes that connect different aspects of the LLM ecosystem. Two, I would say is in engineering. So like distributed training might be one, or just understanding better ways to even be able to train large models, understanding better ways to then distill them or run them. So like this heavy interaction now between engineering and what I might call traditional machine learning from 10 years ago where you had to know a lot of math, you had to know calculus very well, things like that. Now you also need to be, again, a very strong engineer, which is exciting. >> I interviewed Swami when he talked about the news. He's ahead of Amazon's machine learning and AI when they announced Hugging Face announcement. And I reminded him how Amazon was easy to get into if you were developing a startup back in 2007,8, and that the language models had that similar problem. It's step up a lot of content and a lot of expense to get provisioned up, now it's easy. So this is the next wave of innovation. So how do you guys see that from where we are right now? Are we at that point where it's that moment where it's that cloud-like experience for LLMs and large language models? >> Yeah, go ahead John. >> I think the answer is yes. We see a number of large companies that are training these and serving these, some of which are being co-interviewed in this episode. I think we're at that. Like, you can hit one of these with a simple, single line of Python, hitting an API, you can boot this up in seconds if you want. It's easy. >> Got it. >> So I (audio cuts out). >> Well let's take a step back and talk about the company. You guys being featured here on the Showcase. Arthur, what drove you to start the company? How'd this all come together? What's the origination story? Obviously you got a big customers, how'd get started? What are you guys doing? How do you make money? Give a quick overview. >> Yeah, I think John and I come at it from slightly different angles, but for myself, I have been a part of a number of technology companies. I joined Capital One, they acquired my last company and shortly after I joined, they asked me to start their AI team. And so even though I've been doing AI for a long time, I started my career back in DARPA. It was the first time I was really working at scale in AI at an organization where there were hundreds of millions of dollars in revenue at stake with the operation of these models and that they were impacting millions of people's financial livelihoods. And so it just got me hyper-focused on these issues around making sure that your AI worked well and it worked well for your company and it worked well for the people who were being affected by it. At the time when I was doing this 2016, 2017, 2018, there just wasn't any tooling out there to support this production management model monitoring life phase of the life cycle. And so we basically left to start the company that I wanted. And John has a his own story. I'll let let you share that one, John. >> Go ahead John, you're up. >> Yeah, so I'm coming at this from a different world. So I'm on leave now from a tenured role in academia where I was leading a large lab focusing on the intersection of machine learning and economics. And so questions like fairness or the response to the dynamism on the underlying environment have been around for quite a long time in that space. And so I've been thinking very deeply about some of those more like R and D style questions as well as having deployed some automation code across a couple of different industries, some in online advertising, some in the healthcare space and so on, where concerns of, again, fairness come to bear. And so Adam and I connected to understand the space of what that might look like in the 2018 20 19 realm from a quantitative and from a human-centered point of view. And so booted things up from there. >> Yeah, bring that applied engineering R and D into the Capital One, DNA that he had at scale. I could see that fit. I got to ask you now, next step, as you guys move out and think about LLMs and the recent AI news around the generative models and the foundational models like ChatGPT, how should we be looking at that news and everyone watching might be thinking the same thing. I know at the board level companies like, we should refactor our business, this is the future. It's that kind of moment, and the tech team's like, okay, boss, how do we do this again? Or are they prepared? How should we be thinking? How should people watching be thinking about LLMs? >> Yeah, I think they really are transformative. And so, I mean, we're seeing companies all over the place. Everything from large tech companies to a lot of our large enterprise customers are launching significant projects at core parts of their business. And so, yeah, I would be surprised, if you're serious about becoming an AI native company, which most leading companies are, then this is a trend that you need to be taking seriously. And we're seeing the adoption rate. It's funny, I would say the AI adoption in the broader business world really started, let's call it four or five years ago, and it was a relatively slow adoption rate, but I think all that kind of investment in and scaling the maturity curve has paid off because the rate at which people are adopting and deploying systems based on this is tremendous. I mean, this has all just happened in the few months and we're already seeing people get systems into production. So, now there's a lot of things you have to guarantee in order to put these in production in a way that basically is added into your business and doesn't cause more headaches than it solves. And so that's where we help customers is where how do you put these out there in a way that they're going to represent your company well, they're going to perform well, they're going to do their job and do it properly. >> So in the use case, as a customer, as I think about this, there's workflows. They might have had an ML AI ops team that's around IT. Their inference engines are out there. They probably don't have a visibility on say how much it costs, they're kicking the tires. When you look at the deployment, there's a cost piece, there's a workflow piece, there's fairness you mentioned John, what should be, I should be thinking about if I'm going to be deploying stuff into production, I got to think about those things. What's your opinion? >> Yeah, I'm happy to dive in on that one. So monitoring in general is extremely important once you have one of these LLMs in production, and there have been some changes versus traditional monitoring that we can dive deeper into that LLMs are really accelerated. But a lot of that bread and butter style of things you should be looking out for remain just as important as they are for what you might call traditional machine learning models. So the underlying environment of data streams, the way users interact with these models, these are all changing over time. And so any performance metrics that you care about, traditional ones like an accuracy, if you can define that for an LLM, ones around, for example, fairness or bias. If that is a concern for your particular use case and so on. Those need to be tracked. Now there are some interesting changes that LLMs are bringing along as well. So most ML models in production that we see are relatively static in the sense that they're not getting flipped in more than maybe once a day or once a week or they're just set once and then not changed ever again. With LLMs, there's this ongoing value alignment or collection of preferences from users that is often constantly updating the model. And so that opens up all sorts of vectors for, I won't say attack, but for problems to arise in production. Like users might learn to use your system in a different way and thus change the way those preferences are getting collected and thus change your system in ways that you never intended. So maybe that went through governance already internally at the company and now it's totally, totally changed and it's through no fault of your own, but you need to be watching over that for sure. >> Talk about the reinforced learnings from human feedback. How's that factoring in to the LLMs? Is that part of it? Should people be thinking about that? Is that a component that's important? >> It certainly is, yeah. So this is one of the big tweaks that happened with InstructGPT, which is the basis model behind ChatGPT and has since gone on to be used all over the place. So value alignment I think is through RLHF like you mentioned is a very interesting space to get into and it's one that you need to watch over. Like, you're asking humans for feedback over outputs from a model and then you're updating the model with respect to that human feedback. And now you've thrown humans into the loop here in a way that is just going to complicate things. And it certainly helps in many ways. You can ask humans to, let's say that you're deploying an internal chat bot at an enterprise, you could ask humans to align that LLM behind the chatbot to, say company values. And so you're listening feedback about these company values and that's going to scoot that chatbot that you're running internally more toward the kind of language that you'd like to use internally on like a Slack channel or something like that. Watching over that model I think in that specific case, that's a compliance and HR issue as well. So while it is part of the greater LLM stack, you can also view that as an independent bit to watch over. >> Got it, and these are important factors. When people see the Bing news, they freak out how it's doing great. Then it goes off the rails, it goes big, fails big. (laughing) So these models people see that, is that human interaction or is that feedback, is that not accepting it or how do people understand how to take that input in and how to build the right apps around LLMs? This is a tough question. >> Yeah, for sure. So some of the examples that you'll see online where these chatbots go off the rails are obviously humans trying to break the system, but some of them clearly aren't. And that's because these are large statistical models and we don't know what's going to pop out of them all the time. And even if you're doing as much in-house testing at the big companies like the Go-HERE's and the OpenAI's of the world, to try to prevent things like toxicity or racism or other sorts of bad content that might lead to bad pr, you're never going to catch all of these possible holes in the model itself. And so, again, it's very, very important to keep watching over that while it's in production. >> On the business model side, how are you guys doing? What's the approach? How do you guys engage with customers? Take a minute to explain the customer engagement. What do they need? What do you need? How's that work? >> Yeah, I can talk a little bit about that. So it's really easy to get started. It's literally a matter of like just handing out an API key and people can get started. And so we also offer alternative, we also offer versions that can be installed on-prem for models that, we find a lot of our customers have models that deal with very sensitive data. So you can run it in your cloud account or use our cloud version. And so yeah, it's pretty easy to get started with this stuff. We find people start using it a lot of times during the validation phase 'cause that way they can start baselining performance models, they can do champion challenger, they can really kind of baseline the performance of, maybe they're considering different foundation models. And so it's a really helpful tool for understanding differences in the way these models perform. And then from there they can just flow that into their production inferencing, so that as these systems are out there, you have really kind of real time monitoring for anomalies and for all sorts of weird behaviors as well as that continuous feedback loop that helps you make make your product get better and observability and you can run all sorts of aggregated reports to really understand what's going on with these models when they're out there deciding. I should also add that we just today have another way to adopt Arthur and that is we are in the AWS marketplace, and so we are available there just to make it that much easier to use your cloud credits, skip the procurement process, and get up and running really quickly. >> And that's great 'cause Amazon's got SageMaker, which handles a lot of privacy stuff, all kinds of cool things, or you can get down and dirty. So I got to ask on the next one, production is a big deal, getting stuff into production. What have you guys learned that you could share to folks watching? Is there a cost issue? I got to monitor, obviously you brought that up, we talked about the even reinforcement issues, all these things are happening. What is the big learnings that you could share for people that are going to put these into production to watch out for, to plan for, or be prepared for, hope for the best plan for the worst? What's your advice? >> I can give a couple opinions there and I'm sure Adam has. Well, yeah, the big one from my side is, again, I had mentioned this earlier, it's just the input data streams because humans are also exploring how they can use these systems to begin with. It's really, really hard to predict the type of inputs you're going to be seeing in production. Especially, we always talk about chatbots, but then any generative text tasks like this, let's say you're taking in news articles and summarizing them or something like that, it's very hard to get a good sampling even of the set of news articles in such a way that you can really predict what's going to pop out of that model. So to me, it's, adversarial maybe isn't the word that I would use, but it's an unnatural shifting input distribution of like prompts that you might see for these models. That's certainly one. And then the second one that I would talk about is, it can be hard to understand the costs, the inference time costs behind these LLMs. So the pricing on these is always changing as the models change size, it might go up, it might go down based on model size, based on energy cost and so on, but your pricing per token or per a thousand tokens and that I think can be difficult for some clients to wrap their head around. Again, you don't know how these systems are going to be used after all so it can be tough. And so again that's another metric that really should be tracked. >> Yeah, and there's a lot of trade off choices in there with like, how many tokens do you want at each step and in the sequence and based on, you have (indistinct) and you reject these tokens and so based on how your system's operating, that can make the cost highly variable. And that's if you're using like an API version that you're paying per token. A lot of people also choose to run these internally and as John mentioned, the inference time on these is significantly higher than a traditional classifi, even NLP classification model or tabular data model, like orders of magnitude higher. And so you really need to understand how that, as you're constantly iterating on these models and putting out new versions and new features in these models, how that's affecting the overall scale of that inference cost because you can use a lot of computing power very quickly with these profits. >> Yeah, scale, performance, price all come together. I got to ask while we're here on the secret sauce of the company, if you had to describe to people out there watching, what's the secret sauce of the company? What's the key to your success? >> Yeah, so John leads our research team and they've had a number of really cool, I think AI as much as it's been hyped for a while, it's still commercial AI at least is really in its infancy. And so the way we're able to pioneer new ways to think about performance for computer vision NLP LLMs is probably the thing that I'm proudest about. John and his team publish papers all the time at Navs and other places. But I think it's really being able to define what performance means for basically any kind of model type and give people really powerful tools to understand that on an ongoing basis. >> John, secret sauce, how would you describe it? You got all the action happening all around you. >> Yeah, well I going to appreciate Adam talking me up like that. No, I. (all laughing) >> Furrier: Robs to you. >> I would also say a couple of other things here. So we have a very strong engineering team and so I think some early hires there really set the standard at a very high bar that we've maintained as we've grown. And I think that's really paid dividends as scalabilities become even more of a challenge in these spaces, right? And so that's not just scalability when it comes to LLMs, that's scalability when it comes to millions of inferences per day, that kind of thing as well in traditional ML models. And I think that's compared to potential competitors, that's really... Well, it's made us able to just operate more efficiently and pass that along to the client. >> Yeah, and I think the infancy comment is really important because it's the beginning. You really is a long journey ahead. A lot of change coming, like I said, it's a huge wave. So I'm sure you guys got a lot of plannings at the foundation even for your own company, so I appreciate the candid response there. Final question for you guys is, what should the top things be for a company in 2023? If I'm going to set the agenda and I'm a customer moving forward, putting the pedal to the metal, so to speak, what are the top things I should be prioritizing or I need to do to be successful with AI in 2023? >> Yeah, I think, so number one, as we talked about, we've been talking about this entire episode, the things are changing so quickly and the opportunities for business transformation and really disrupting different applications, different use cases, is almost, I don't think we've even fully comprehended how big it is. And so really digging in to your business and understanding where I can apply these new sets of foundation models is, that's a top priority. The interesting thing is I think there's another force at play, which is the macroeconomic conditions and a lot of places are, they're having to work harder to justify budgets. So in the past, couple years ago maybe, they had a blank check to spend on AI and AI development at a lot of large enterprises that was limited primarily by the amount of talent they could scoop up. Nowadays these expenditures are getting scrutinized more. And so one of the things that we really help our customers with is like really calculating the ROI on these things. And so if you have models out there performing and you have a new version that you can put out that lifts the performance by 3%, how many tens of millions of dollars does that mean in business benefit? Or if I want to go to get approval from the CFO to spend a few million dollars on this new project, how can I bake in from the beginning the tools to really show the ROI along the way? Because I think in these systems when done well for a software project, the ROI can be like pretty spectacular. Like we see over a hundred percent ROI in the first year on some of these projects. And so, I think in 2023, you just need to be able to show what you're getting for that spend. >> It's a needle moving moment. You see it all the time with some of these aha moments or like, whoa, blown away. John, I want to get your thoughts on this because one of the things that comes up a lot for companies that I talked to, that are on my second wave, I would say coming in, maybe not, maybe the front wave of adopters is talent and team building. You mentioned some of the hires you got were game changing for you guys and set the bar high. As you move the needle, new developers going to need to come in. What's your advice given that you've been a professor, you've seen students, I know a lot of computer science people want to shift, they might not be yet skilled in AI, but they're proficient in programming, is that's going to be another opportunity with open source when things are happening. How do you talk to that next level of talent that wants to come in to this market to supplement teams and be on teams, lead teams? Any advice you have for people who want to build their teams and people who are out there and want to be a coder in AI? >> Yeah, I've advice, and this actually works for what it would take to be a successful AI company in 2023 as well, which is, just don't be afraid to iterate really quickly with these tools. The space is still being explored on what they can be used for. A lot of the tasks that they're used for now right? like creating marketing content using a machine learning is not a new thing to do. It just works really well now. And so I'm excited to see what the next year brings in terms of folks from outside of core computer science who are, other engineers or physicists or chemists or whatever who are learning how to use these increasingly easy to use tools to leverage LLMs for tasks that I think none of us have really thought about before. So that's really, really exciting. And so toward that I would say iterate quickly. Build things on your own, build demos, show them the friends, host them online and you'll learn along the way and you'll have somebody to show for it. And also you'll help us explore that space. >> Guys, congratulations with Arthur. Great company, great picks and shovels opportunities out there for everybody. Iterate fast, get in quickly and don't be afraid to iterate. Great advice and thank you for coming on and being part of the AWS showcase, thanks. >> Yeah, thanks for having us on John. Always a pleasure. >> Yeah, great stuff. Adam Wenchel, John Dickerson with Arthur. Thanks for coming on theCUBE. I'm John Furrier, your host. Generative AI and AWS. Keep it right there for more action with theCUBE. Thanks for watching. (upbeat music)

(upbeat music) >> Hello, everyone, and welcome to theCUBE's presentation of the "AWS Startup Showcase." This is season three, episode one. The focus of this episode is AI/ML: Top Startups Building Foundational Models, Infrastructure, and AI. It's great topics, super-relevant, and it's part of our ongoing coverage of startups in the AWS ecosystem. I'm your host, John Furrier, with theCUBE. Today, we're excited to be joined by Jay Marshall, VP of Business Development at Neural Magic. Jay, thanks for coming on theCUBE. >> Hey, John, thanks so much. Thanks for having us. >> We had a great CUBE conversation with you guys. This is very much about the company focuses. It's a feature presentation for the "Startup Showcase," and the machine learning at scale is the topic, but in general, it's more, (laughs) and we should call it "Machine Learning and AI: How to Get Started," because everybody is retooling their business. Companies that aren't retooling their business right now with AI first will be out of business, in my opinion. You're seeing massive shift. This is really truly the beginning of the next-gen machine learning AI trend. It's really seeing ChatGPT. Everyone sees that. That went mainstream. But this is just the beginning. This is scratching the surface of this next-generation AI with machine learning powering it, and with all the goodness of cloud, cloud scale, and how horizontally scalable it is. The resources are there. You got the Edge. Everything's perfect for AI 'cause data infrastructure's exploding in value. AI is just the applications. This is a super topic, so what do you guys see in this general area of opportunities right now in the headlines? And I'm sure you guys' phone must be ringing off the hook, metaphorically speaking, or emails and meetings and Zooms. What's going on over there at Neural Magic? >> No, absolutely, and you pretty much nailed most of it. I think that, you know, my background, we've seen for the last 20-plus years. Even just getting enterprise applications kind of built and delivered at scale, obviously, amazing things with AWS and the cloud to help accelerate that. And we just kind of figured out in the last five or so years how to do that productively and efficiently, kind of from an operations perspective. Got development and operations teams. We even came up with DevOps, right? But now, we kind of have this new kind of persona and new workload that developers have to talk to, and then it has to be deployed on those ITOps solutions. And so you pretty much nailed it. Folks are saying, "Well, how do I do this?" These big, generational models or foundational models, as we're calling them, they're great, but enterprises want to do that with their data, on their infrastructure, at scale, at the edge. So for us, yeah, we're helping enterprises accelerate that through optimizing models and then delivering them at scale in a more cost-effective fashion. >> Yeah, and I think one of the things, the benefits of OpenAI we saw, was not only is it open source, then you got also other models that are more proprietary, is that it shows the world that this is really happening, right? It's a whole nother level, and there's also new landscape kind of maps coming out. You got the generative AI, and you got the foundational models, large LLMs. Where do you guys fit into the landscape? Because you guys are in the middle of this. How do you talk to customers when they say, "I'm going down this road. I need help. I'm going to stand this up." This new AI infrastructure and applications, where do you guys fit in the landscape? >> Right, and really, the answer is both. I think today, when it comes to a lot of what for some folks would still be considered kind of cutting edge around computer vision and natural language processing, a lot of our optimization tools and our runtime are based around most of the common computer vision and natural language processing models. So your YOLOs, your BERTs, you know, your DistilBERTs and what have you, so we work to help optimize those, again, who've gotten great performance and great value for customers trying to get those into production. But when you get into the LLMs, and you mentioned some of the open source components there, our research teams have kind of been right in the trenches with those. So kind of the GPT open source equivalent being OPT, being able to actually take, you know, a multi-$100 billion parameter model and sparsify that or optimize that down, shaving away a ton of parameters, and being able to run it on smaller infrastructure. So I think the evolution here, you know, all this stuff came out in the last six months in terms of being turned loose into the wild, but we're staying in the trenches with folks so that we can help optimize those as well and not require, again, the heavy compute, the heavy cost, the heavy power consumption as those models evolve as well. So we're staying right in with everybody while they're being built, but trying to get folks into production today with things that help with business value today. >> Jay, I really appreciate you coming on theCUBE, and before we came on camera, you said you just were on a customer call. I know you got a lot of activity. What specific things are you helping enterprises solve? What kind of problems? Take us through the spectrum from the beginning, people jumping in the deep end of the pool, some people kind of coming in, starting out slow. What are the scale? Can you scope the kind of use cases and problems that are emerging that people are calling you for? >> Absolutely, so I think if I break it down to kind of, like, your startup, or I maybe call 'em AI native to kind of steal from cloud native years ago, that group, it's pretty much, you know, part and parcel for how that group already runs. So if you have a data science team and an ML engineering team, you're building models, you're training models, you're deploying models. You're seeing firsthand the expense of starting to try to do that at scale. So it's really just a pure operational efficiency play. They kind of speak natively to our tools, which we're doing in the open source. So it's really helping, again, with the optimization of the models they've built, and then, again, giving them an alternative to expensive proprietary hardware accelerators to have to run them. Now, on the enterprise side, it varies, right? You have some kind of AI native folks there that already have these teams, but you also have kind of, like, AI curious, right? Like, they want to do it, but they don't really know where to start, and so for there, we actually have an open source toolkit that can help you get into this optimization, and then again, that runtime, that inferencing runtime, purpose-built for CPUs. It allows you to not have to worry, again, about do I have a hardware accelerator available? How do I integrate that into my application stack? If I don't already know how to build this into my infrastructure, does my ITOps teams, do they know how to do this, and what does that runway look like? How do I cost for this? How do I plan for this? When it's just x86 compute, we've been doing that for a while, right? So it obviously still requires more, but at least it's a little bit more predictable. >> It's funny you mentioned AI native. You know, born in the cloud was a phrase that was out there. Now, you have startups that are born in AI companies. So I think you have this kind of cloud kind of vibe going on. You have lift and shift was a big discussion. Then you had cloud native, kind of in the cloud, kind of making it all work. Is there a existing set of things? People will throw on this hat, and then what's the difference between AI native and kind of providing it to existing stuff? 'Cause we're a lot of people take some of these tools and apply it to either existing stuff almost, and it's not really a lift and shift, but it's kind of like bolting on AI to something else, and then starting with AI first or native AI. >> Absolutely. It's a- >> How would you- >> It's a great question. I think that probably, where I'd probably pull back to kind of allow kind of retail-type scenarios where, you know, for five, seven, nine years or more even, a lot of these folks already have data science teams, you know? I mean, they've been doing this for quite some time. The difference is the introduction of these neural networks and deep learning, right? Those kinds of models are just a little bit of a paradigm shift. So, you know, I obviously was trying to be fun with the term AI native, but I think it's more folks that kind of came up in that neural network world, so it's a little bit more second nature, whereas I think for maybe some traditional data scientists starting to get into neural networks, you have the complexity there and the training overhead, and a lot of the aspects of getting a model finely tuned and hyperparameterization and all of these aspects of it. It just adds a layer of complexity that they're just not as used to dealing with. And so our goal is to help make that easy, and then of course, make it easier to run anywhere that you have just kind of standard infrastructure. >> Well, the other point I'd bring out, and I'd love to get your reaction to, is not only is that a neural network team, people who have been focused on that, but also, if you look at some of the DataOps lately, AIOps markets, a lot of data engineering, a lot of scale, folks who have been kind of, like, in that data tsunami cloud world are seeing, they kind of been in this, right? They're, like, been experiencing that. >> No doubt. I think it's funny the data lake concept, right? And you got data oceans now. Like, the metaphors just keep growing on us, but where it is valuable in terms of trying to shift the mindset, I've always kind of been a fan of some of the naming shift. I know with AWS, they always talk about purpose-built databases. And I always liked that because, you know, you don't have one database that can do everything. Even ones that say they can, like, you still have to do implementation detail differences. So sitting back and saying, "What is my use case, and then which database will I use it for?" I think it's kind of similar here. And when you're building those data teams, if you don't have folks that are doing data engineering, kind of that data harvesting, free processing, you got to do all that before a model's even going to care about it. So yeah, it's definitely a central piece of this as well, and again, whether or not you're going to be AI negative as you're making your way to kind of, you know, on that journey, you know, data's definitely a huge component of it. >> Yeah, you would have loved our Supercloud event we had. Talk about naming and, you know, around data meshes was talked about a lot. You're starting to see the control plane layers of data. I think that was the beginning of what I saw as that data infrastructure shift, to be horizontally scalable. So I have to ask you, with Neural Magic, when your customers and the people that are prospects for you guys, they're probably asking a lot of questions because I think the general thing that we see is, "How do I get started? Which GPU do I use?" I mean, there's a lot of things that are kind of, I won't say technical or targeted towards people who are living in that world, but, like, as the mainstream enterprises come in, they're going to need a playbook. What do you guys see, what do you guys offer your clients when they come in, and what do you recommend? >> Absolutely, and I think where we hook in specifically tends to be on the training side. So again, I've built a model. Now, I want to really optimize that model. And then on the runtime side when you want to deploy it, you know, we run that optimized model. And so that's where we're able to provide. We even have a labs offering in terms of being able to pair up our engineering teams with a customer's engineering teams, and we can actually help with most of that pipeline. So even if it is something where you have a dataset and you want some help in picking a model, you want some help training it, you want some help deploying that, we can actually help there as well. You know, there's also a great partner ecosystem out there, like a lot of folks even in the "Startup Showcase" here, that extend beyond into kind of your earlier comment around data engineering or downstream ITOps or the all-up MLOps umbrella. So we can absolutely engage with our labs, and then, of course, you know, again, partners, which are always kind of key to this. So you are spot on. I think what's happened with the kind of this, they talk about a hockey stick. This is almost like a flat wall now with the rate of innovation right now in this space. And so we do have a lot of folks wanting to go straight from curious to native. And so that's definitely where the partner ecosystem comes in so hard 'cause there just isn't anybody or any teams out there that, I literally do from, "Here's my blank database, and I want an API that does all the stuff," right? Like, that's a big chunk, but we can definitely help with the model to delivery piece. >> Well, you guys are obviously a featured company in this space. Talk about the expertise. A lot of companies are like, I won't say faking it till they make it. You can't really fake security. You can't really fake AI, right? So there's going to be a learning curve. They'll be a few startups who'll come out of the gate early. You guys are one of 'em. Talk about what you guys have as expertise as a company, why you're successful, and what problems do you solve for customers? >> No, appreciate that. Yeah, we actually, we love to tell the story of our founder, Nir Shavit. So he's a 20-year professor at MIT. Actually, he was doing a lot of work on kind of multicore processing before there were even physical multicores, and actually even did a stint in computational neurobiology in the 2010s, and the impetus for this whole technology, has a great talk on YouTube about it, where he talks about the fact that his work there, he kind of realized that the way neural networks encode and how they're executed by kind of ramming data layer by layer through these kind of HPC-style platforms, actually was not analogous to how the human brain actually works. So we're on one side, we're building neural networks, and we're trying to emulate neurons. We're not really executing them that way. So our team, which one of the co-founders, also an ex-MIT, that was kind of the birth of why can't we leverage this super-performance CPU platform, which has those really fat, fast caches attached to each core, and actually start to find a way to break that model down in a way that I can execute things in parallel, not having to do them sequentially? So it is a lot of amazing, like, talks and stuff that show kind of the magic, if you will, a part of the pun of Neural Magic, but that's kind of the foundational layer of all the engineering that we do here. And in terms of how we're able to bring it to reality for customers, I'll give one customer quote where it's a large retailer, and it's a people-counting application. So a very common application. And that customer's actually been able to show literally double the amount of cameras being run with the same amount of compute. So for a one-to-one perspective, two-to-one, business leaders usually like that math, right? So we're able to show pure cost savings, but even performance-wise, you know, we have some of the common models like your ResNets and your YOLOs, where we can actually even perform better than hardware-accelerated solutions. So we're trying to do, I need to just dumb it down to better, faster, cheaper, but from a commodity perspective, that's where we're accelerating. >> That's not a bad business model. Make things easier to use, faster, and reduce the steps it takes to do stuff. So, you know, that's always going to be a good market. Now, you guys have DeepSparse, which we've talked about on our CUBE conversation prior to this interview, delivers ML models through the software so the hardware allows for a decoupling, right? >> Yep. >> Which is going to drive probably a cost advantage. Also, it's also probably from a deployment standpoint it must be easier. Can you share the benefits? Is it a cost side? Is it more of a deployment? What are the benefits of the DeepSparse when you guys decouple the software from the hardware on the ML models? >> No you actually, you hit 'em both 'cause that really is primarily the value. Because ultimately, again, we're so early. And I came from this world in a prior life where I'm doing Java development, WebSphere, WebLogic, Tomcat open source, right? When we were trying to do innovation, we had innovation buckets, 'cause everybody wanted to be on the web and have their app and a browser, right? We got all the money we needed to build something and show, hey, look at the thing on the web, right? But when you had to get in production, that was the challenge. So to what you're speaking to here, in this situation, we're able to show we're just a Python package. So whether you just install it on the operating system itself, or we also have a containerized version you can drop on any container orchestration platform, so ECS or EKS on AWS. And so you get all the auto-scaling features. So when you think about that kind of a world where you have everything from real-time inferencing to kind of after hours batch processing inferencing, the fact that you can auto scale that hardware up and down and it's CPU based, so you're paying by the minute instead of maybe paying by the hour at a lower cost shelf, it does everything from pure cost to, again, I can have my standard IT team say, "Hey, here's the Kubernetes in the container," and it just runs on the infrastructure we're already managing. So yeah, operational, cost and again, and many times even performance. (audio warbles) CPUs if I want to. >> Yeah, so that's easier on the deployment too. And you don't have this kind of, you know, blank check kind of situation where you don't know what's on the backend on the cost side. >> Exactly. >> And you control the actual hardware and you can manage that supply chain. >> And keep in mind, exactly. Because the other thing that sometimes gets lost in the conversation, depending on where a customer is, some of these workloads, like, you know, you and I remember a world where even like the roundtrip to the cloud and back was a problem for folks, right? We're used to extremely low latency. And some of these workloads absolutely also adhere to that. But there's some workloads where the latency isn't as important. And we actually even provide the tuning. Now, if we're giving you five milliseconds of latency and you don't need that, you can tune that back. So less CPU, lower cost. Now, throughput and other things come into play. But that's the kind of configurability and flexibility we give for operations. >> All right, so why should I call you if I'm a customer or prospect Neural Magic, what problem do I have or when do I know I need you guys? When do I call you in and what does my environment look like? When do I know? What are some of the signals that would tell me that I need Neural Magic? >> No, absolutely. So I think in general, any neural network, you know, the process I mentioned before called sparcification, it's, you know, an optimization process that we specialize in. Any neural network, you know, can be sparcified. So I think if it's a deep-learning neural network type model. If you're trying to get AI into production, you have cost concerns even performance-wise. I certainly hate to be too generic and say, "Hey, we'll talk to everybody." But really in this world right now, if it's a neural network, it's something where you're trying to get into production, you know, we are definitely offering, you know, kind of an at-scale performant deployable solution for deep learning models. >> So neural network you would define as what? Just devices that are connected that need to know about each other? What's the state-of-the-art current definition of neural network for customers that may think they have a neural network or might not know they have a neural network architecture? What is that definition for neural network? >> That's a great question. So basically, machine learning models that fall under this kind of category, you hear about transformers a lot, or I mentioned about YOLO, the YOLO family of computer vision models, or natural language processing models like BERT. If you have a data science team or even developers, some even regular, I used to call myself a nine to five developer 'cause I worked in the enterprise, right? So like, hey, we found a new open source framework, you know, I used to use Spring back in the day and I had to go figure it out. There's developers that are pulling these models down and they're figuring out how to get 'em into production, okay? So I think all of those kinds of situations, you know, if it's a machine learning model of the deep learning variety that's, you know, really specifically where we shine. >> Okay, so let me pretend I'm a customer for a minute. I have all these videos, like all these transcripts, I have all these people that we've interviewed, CUBE alumnis, and I say to my team, "Let's AI-ify, sparcify theCUBE." >> Yep. >> What do I do? I mean, do I just like, my developers got to get involved and they're going to be like, "Well, how do I upload it to the cloud? Do I use a GPU?" So there's a thought process. And I think a lot of companies are going through that example of let's get on this AI, how can it help our business? >> Absolutely. >> What does that progression look like? Take me through that example. I mean, I made up theCUBE example up, but we do have a lot of data. We have large data models and we have people and connect to the internet and so we kind of seem like there's a neural network. I think every company might have a neural network in place. >> Well, and I was going to say, I think in general, you all probably do represent even the standard enterprise more than most. 'Cause even the enterprise is going to have a ton of video content, a ton of text content. So I think it's a great example. So I think that that kind of sea or I'll even go ahead and use that term data lake again, of data that you have, you're probably going to want to be setting up kind of machine learning pipelines that are going to be doing all of the pre-processing from kind of the raw data to kind of prepare it into the format that say a YOLO would actually use or let's say BERT for natural language processing. So you have all these transcripts, right? So we would do a pre-processing path where we would create that into the file format that BERT, the machine learning model would know how to train off of. So that's kind of all the pre-processing steps. And then for training itself, we actually enable what's called sparse transfer learning. So that's transfer learning is a very popular method of doing training with existing models. So we would be able to retrain that BERT model with your transcript data that we have now done the pre-processing with to get it into the proper format. And now we have a BERT natural language processing model that's been trained on your data. And now we can deploy that onto DeepSparse runtime so that now you can ask that model whatever questions, or I should say pass, you're not going to ask it those kinds of questions ChatGPT, although we can do that too. But you're going to pass text through the BERT model and it's going to give you answers back. It could be things like sentiment analysis or text classification. You just call the model, and now when you pass text through it, you get the answers better, faster or cheaper. I'll use that reference again. >> Okay, we can create a CUBE bot to give us questions on the fly from the the AI bot, you know, from our previous guests. >> Well, and I will tell you using that as an example. So I had mentioned OPT before, kind of the open source version of ChatGPT. So, you know, typically that requires multiple GPUs to run. So our research team, I may have mentioned earlier, we've been able to sparcify that over 50% already and run it on only a single GPU. And so in that situation, you could train OPT with that corpus of data and do exactly what you say. Actually we could use Alexa, we could use Alexa to actually respond back with voice. How about that? We'll do an API call and we'll actually have an interactive Alexa-enabled bot. >> Okay, we're going to be a customer, let's put it on the list. But this is a great example of what you guys call software delivered AI, a topic we chatted about on theCUBE conversation. This really means this is a developer opportunity. This really is the convergence of the data growth, the restructuring, how data is going to be horizontally scalable, meets developers. So this is an AI developer model going on right now, which is kind of unique. >> It is, John, I will tell you what's interesting. And again, folks don't always think of it this way, you know, the AI magical goodness is now getting pushed in the middle where the developers and IT are operating. And so it again, that paradigm, although for some folks seem obvious, again, if you've been around for 20 years, that whole all that plumbing is a thing, right? And so what we basically help with is when you deploy the DeepSparse runtime, we have a very rich API footprint. And so the developers can call the API, ITOps can run it, or to your point, it's developer friendly enough that you could actually deploy our off-the-shelf models. We have something called the SparseZoo where we actually publish pre-optimized or pre-sparcified models. And so developers could literally grab those right off the shelf with the training they've already had and just put 'em right into their applications and deploy them as containers. So yeah, we enable that for sure as well. >> It's interesting, DevOps was infrastructure as code and we had a last season, a series on data as code, which we kind of coined. This is data as code. This is a whole nother level of opportunity where developers just want to have programmable data and apps with AI. This is a whole new- >> Absolutely. >> Well, absolutely great, great stuff. Our news team at SiliconANGLE and theCUBE said you guys had a little bit of a launch announcement you wanted to make here on the "AWS Startup Showcase." So Jay, you have something that you want to launch here? >> Yes, and thank you John for teeing me up. So I'm going to try to put this in like, you know, the vein of like an AWS, like main stage keynote launch, okay? So we're going to try this out. So, you know, a lot of our product has obviously been built on top of x86. I've been sharing that the past 15 minutes or so. And with that, you know, we're seeing a lot of acceleration for folks wanting to run on commodity infrastructure. But we've had customers and prospects and partners tell us that, you know, ARM and all of its kind of variance are very compelling, both cost performance-wise and also obviously with Edge. And wanted to know if there was anything we could do from a runtime perspective with ARM. And so we got the work and, you know, it's a hard problem to solve 'cause the instructions set for ARM is very different than the instruction set for x86, and our deep tensor column technology has to be able to work with that lower level instruction spec. But working really hard, the engineering team's been at it and we are happy to announce here at the "AWS Startup Showcase," that DeepSparse inference now has, or inference runtime now has support for AWS Graviton instances. So it's no longer just x86, it is also ARM and that obviously also opens up the door to Edge and further out the stack so that optimize once run anywhere, we're not going to open up. So it is an early access. So if you go to neuralmagic.com/graviton, you can sign up for early access, but we're excited to now get into the ARM side of the fence as well on top of Graviton. >> That's awesome. Our news team is going to jump on that news. We'll get it right up. We get a little scoop here on the "Startup Showcase." Jay Marshall, great job. That really highlights the flexibility that you guys have when you decouple the software from the hardware. And again, we're seeing open source driving a lot more in AI ops now with with machine learning and AI. So to me, that makes a lot of sense. And congratulations on that announcement. Final minute or so we have left, give a summary of what you guys are all about. Put a plug in for the company, what you guys are looking to do. I'm sure you're probably hiring like crazy. Take the last few minutes to give a plug for the company and give a summary. >> No, I appreciate that so much. So yeah, joining us out neuralmagic.com, you know, part of what we didn't spend a lot of time here, our optimization tools, we are doing all of that in the open source. It's called SparseML and I mentioned SparseZoo briefly. So we really want the data scientists community and ML engineering community to join us out there. And again, the DeepSparse runtime, it's actually free to use for trial purposes and for personal use. So you can actually run all this on your own laptop or on an AWS instance of your choice. We are now live in the AWS marketplace. So push button, deploy, come try us out and reach out to us on neuralmagic.com. And again, sign up for the Graviton early access. >> All right, Jay Marshall, Vice President of Business Development Neural Magic here, talking about performant, cost effective machine learning at scale. This is season three, episode one, focusing on foundational models as far as building data infrastructure and AI, AI native. I'm John Furrier with theCUBE. Thanks for watching. (bright upbeat music)

(soft music) >> Hello, everyone. Welcome to theCUBE's presentation of the AWS Startup Showcase. AI and Machine Learning: Top Startups Building Foundational Model Infrastructure. This is season 3, episode 1 of the ongoing series covering the exciting stuff from the AWS ecosystem, talking about machine learning and AI. I'm your host, John Furrier and today we are excited to be joined by Luis Ceze who's the CEO of OctoML and Anna Connolly, VP of customer success and experience OctoML. Great to have you on again, Luis. Anna, thanks for coming on. Appreciate it. >> Thank you, John. It's great to be here. >> Thanks for having us. >> I love the company. We had a CUBE conversation about this. You guys are really addressing how to run foundational models faster for less. And this is like the key theme. But before we get into it, this is a hot trend, but let's explain what you guys do. Can you set the narrative of what the company's about, why it was founded, what's your North Star and your mission? >> Yeah, so John, our mission is to make AI sustainable and accessible for everyone. And what we offer customers is, you know, a way of taking their models into production in the most efficient way possible by automating the process of getting a model and optimizing it for a variety of hardware and making cost-effective. So better, faster, cheaper model deployment. >> You know, the big trend here is AI. Everyone's seeing the ChatGPT, kind of the shot heard around the world. The BingAI and this fiasco and the ongoing experimentation. People are into it, and I think the business impact is clear. I haven't seen this in all of my career in the technology industry of this kind of inflection point. And every senior leader I talk to is rethinking about how to rebuild their business with AI because now the large language models have come in, these foundational models are here, they can see value in their data. This is a 10 year journey in the big data world. Now it's impacting that, and everyone's rebuilding their company around this idea of being AI first 'cause they see ways to eliminate things and make things more efficient. And so now they telling 'em to go do it. And they're like, what do we do? So what do you guys think? Can you explain what is this wave of AI and why is it happening, why now, and what should people pay attention to? What does it mean to them? >> Yeah, I mean, it's pretty clear by now that AI can do amazing things that captures people's imaginations. And also now can show things that are really impactful in businesses, right? So what people have the opportunity to do today is to either train their own model that adds value to their business or find open models out there that can do very valuable things to them. So the next step really is how do you take that model and put it into production in a cost-effective way so that the business can actually get value out of it, right? >> Anna, what's your take? Because customers are there, you're there to make 'em successful, you got the new secret weapon for their business. >> Yeah, I think we just see a lot of companies struggle to get from a trained model into a model that is deployed in a cost-effective way that actually makes sense for the application they're building. I think that's a huge challenge we see today, kind of across the board across all of our customers. >> Well, I see this, everyone asking the same question. I have data, I want to get value out of it. I got to get these big models, I got to train it. What's it going to cost? So I think there's a reality of, okay, I got to do it. Then no one has any visibility on what it costs. When they get into it, this is going to break the bank. So I have to ask you guys, the cost of training these models is on everyone's mind. OctoML, your company's focus on the cost side of it as well as the efficiency side of running these models in production. Why are the production costs such a concern and where specifically are people looking at it and why did it get here? >> Yeah, so training costs get a lot of attention because normally a large number, but we shouldn't forget that it's a large, typically one time upfront cost that customers pay. But, you know, when the model is put into production, the cost grows directly with model usage and you actually want your model to be used because it's adding value, right? So, you know, the question that a customer faces is, you know, they have a model, they have a trained model and now what? So how much would it cost to run in production, right? And now without the big wave in generative AI, which rightfully is getting a lot of attention because of the amazing things that it can do. It's important for us to keep in mind that generative AI models like ChatGPT are huge, expensive energy hogs. They cost a lot to run, right? And given that model usage growth directly, model cost grows directly with usage, what you want to do is make sure that once you put a model into production, you have the best cost structure possible so that you're not surprised when it's gets popular, right? So let me give you an example. So if you have a model that costs, say 1 to $2 million to train, but then it costs about one to two cents per session to use it, right? So if you have a million active users, even if they use just once a day, it's 10 to $20,000 a day to operate that model in production. And that very, very quickly, you know, get beyond what you paid to train it. >> Anna, these aren't small numbers, and it's cost to train and cost to operate, it kind of reminds me of when the cloud came around and the data center versus cloud options. Like, wait a minute, one, it costs a ton of cash to deploy, and then running it. This is kind of a similar dynamic. What are you seeing? >> Yeah, absolutely. I think we are going to see increasingly the cost and production outpacing the costs and training by a lot. I mean, people talk about training costs now because that's what they're confronting now because people are so focused on getting models performant enough to even use in an application. And now that we have them and they're that capable, we're really going to start to see production costs go up a lot. >> Yeah, Luis, if you don't mind, I know this might be a little bit of a tangent, but, you know, training's super important. I get that. That's what people are doing now, but then there's the deployment side of production. Where do people get caught up and miss the boat or misconfigure? What's the gotcha? Where's the trip wire or so to speak? Where do people mess up on the cost side? What do they do? Is it they don't think about it, they tie it to proprietary hardware? What's the issue? >> Yeah, several things, right? So without getting really technical, which, you know, I might get into, you know, you have to understand relationship between performance, you know, both in terms of latency and throughput and cost, right? So reducing latency is important because you improve responsiveness of the model. But it's really important to keep in mind that it often leads diminishing returns. Below a certain latency, making it faster won't make a measurable difference in experience, but it's going to cost a lot more. So understanding that is important. Now, if you care more about throughputs, which is the time it takes for you to, you know, units per period of time, you care about time to solution, we should think about this throughput per dollar. And understand what you want is the highest throughput per dollar, which may come at the cost of higher latency, which you're not going to care about, right? So, and the reality here, John, is that, you know, humans and especially folks in this space want to have the latest and greatest hardware. And often they commit a lot of money to get access to them and have to commit upfront before they understand the needs that their models have, right? So common mistake here, one is not spending time to understand what you really need, and then two, over-committing and using more hardware than you actually need. And not giving yourself enough freedom to get your workload to move around to the more cost-effective choice, right? So this is just a metaphoric choice. And then another thing that's important here too is making a model run faster on the hardware directly translates to lower cost, right? So, but it takes a lot of engineers, you need to think of ways of producing very efficient versions of your model for the target hardware that you're going to use. >> Anna, what's the customer angle here? Because price performance has been around for a long time, people get that, but now latency and throughput, that's key because we're starting to see this in apps. I mean, there's an end user piece. I even seeing it on the infrastructure side where they're taking a heavy lifting away from operational costs. So you got, you know, application specific to the user and/or top of the stack, and then you got actually being used in operations where they want both. >> Yeah, absolutely. Maybe I can illustrate this with a quick story with the customer that we had recently been working with. So this customer is planning to run kind of a transformer based model for tech generation at super high scale on Nvidia T4 GPU, so kind of a commodity GPU. And the scale was so high that they would've been paying hundreds of thousands of dollars in cloud costs per year just to serve this model alone. You know, one of many models in their application stack. So we worked with this team to optimize our model and then benchmark across several possible targets. So that matching the hardware that Luis was just talking about, including the newer kind of Nvidia A10 GPUs. And what they found during this process was pretty interesting. First, the team was able to shave a quarter of their spend just by using better optimization techniques on the T4, the older hardware. But actually moving to a newer GPU would allow them to serve this model in a sub two milliseconds latency, so super fast, which was able to unlock an entirely new kind of user experience. So they were able to kind of change the value they're delivering in their application just because they were able to move to this new hardware easily. So they ultimately decided to plan their deployment on the more expensive A10 because of this, but because of the hardware specific optimizations that we helped them with, they managed to even, you know, bring costs down from what they had originally planned. And so if you extend this kind of example to everything that's happening with generative AI, I think the story we just talked about was super relevant, but the scale can be even higher, you know, it can be tenfold that. We were recently conducting kind of this internal study using GPT-J as a proxy to illustrate the experience of just a company trying to use one of these large language models with an example scenario of creating a chatbot to help job seekers prepare for interviews. So if you imagine kind of a conservative usage scenario where the model generates just 3000 words per user per day, which is, you know, pretty conservative for how people are interacting with these models. It costs 5 cents a session and if you're a company and your app goes viral, so from, you know, beginning of the year there's nobody, at the end of the year there's a million daily active active users in that year alone, going from zero to a million. You'll be spending about $6 million a year, which is pretty unmanageable. That's crazy, right? >> Yeah. >> For a company or a product that's just launching. So I think, you know, for us we see the real way to make these kind of advancements accessible and sustainable, as we said is to bring down cost to serve using these techniques. >> That's a great story and I think that illustrates this idea that deployment cost can vary from situation to situation, from model to model and that the efficiency is so strong with this new wave, it eliminates heavy lifting, creates more efficiency, automates intellect. I mean, this is the trend, this is radical, this is going to increase. So the cost could go from nominal to millions, literally, potentially. So, this is what customers are doing. Yeah, that's a great story. What makes sense on a financial, is there a cost of ownership? Is there a pattern for best practice for training? What do you guys advise cuz this is a lot of time and money involved in all potential, you know, good scenarios of upside. But you can get over your skis as they say, and be successful and be out of business if you don't manage it. I mean, that's what people are talking about, right? >> Yeah, absolutely. I think, you know, we see kind of three main vectors to reduce cost. I think one is make your deployment process easier overall, so that your engineering effort to even get your app running goes down. Two, would be get more from the compute you're already paying for, you're already paying, you know, for your instances in the cloud, but can you do more with that? And then three would be shop around for lower cost hardware to match your use case. So on the first one, I think making the deployment easier overall, there's a lot of manual work that goes into benchmarking, optimizing and packaging models for deployment. And because the performance of machine learning models can be really hardware dependent, you have to go through this process for each target you want to consider running your model on. And this is hard, you know, we see that every day. But for teams who want to incorporate some of these large language models into their applications, it might be desirable because licensing a model from a large vendor like OpenAI can leave you, you know, over provision, kind of paying for capabilities you don't need in your application or can lock you into them and you lose flexibility. So we have a customer whose team actually prepares models for deployment in a SaaS application that many of us use every day. And they told us recently that without kind of an automated benchmarking and experimentation platform, they were spending several days each to benchmark a single model on a single hardware type. So this is really, you know, manually intensive and then getting more from the compute you're already paying for. We do see customers who leave money on the table by running models that haven't been optimized specifically for the hardware target they're using, like Luis was mentioning. And for some teams they just don't have the time to go through an optimization process and for others they might lack kind of specialized expertise and this is something we can bring. And then on shopping around for different hardware types, we really see a huge variation in model performance across hardware, not just CPU vs. GPU, which is, you know, what people normally think of. But across CPU vendors themselves, high memory instances and across cloud providers even. So the best strategy here is for teams to really be able to, we say, look before you leap by running real world benchmarking and not just simulations or predictions to find the best software, hardware combination for their workload. >> Yeah. You guys sound like you have a very impressive customer base deploying large language models. Where would you categorize your current customer base? And as you look out, as you guys are growing, you have new customers coming in, take me through the progression. Take me through the profile of some of your customers you have now, size, are they hyperscalers, are they big app folks, are they kicking the tires? And then as people are out there scratching heads, I got to get in this game, what's their psychology like? Are they coming in with specific problems or do they have specific orientation point of view about what they want to do? Can you share some data around what you're seeing? >> Yeah, I think, you know, we have customers that kind of range across the spectrum of sophistication from teams that basically don't have MLOps expertise in their company at all. And so they're really looking for us to kind of give a full service, how should I do everything from, you know, optimization, find the hardware, prepare for deployment. And then we have teams that, you know, maybe already have their serving and hosting infrastructure up and ready and they already have models in production and they're really just looking to, you know, take the extra juice out of the hardware and just do really specific on that optimization piece. I think one place where we're doing a lot more work now is kind of in the developer tooling, you know, model selection space. And that's kind of an area that we're creating more tools for, particularly within the PyTorch ecosystem to bring kind of this power earlier in the development cycle so that as people are grabbing a model off the shelf, they can, you know, see how it might perform and use that to inform their development process. >> Luis, what's the big, I like this idea of picking the models because isn't that like going to the market and picking the best model for your data? It's like, you know, it's like, isn't there a certain approaches? What's your view on this? 'Cause this is where everyone, I think it's going to be a land rush for this and I want to get your thoughts. >> For sure, yeah. So, you know, I guess I'll start with saying the one main takeaway that we got from the GPT-J study is that, you know, having a different understanding of what your model's compute and memory requirements are, very quickly, early on helps with the much smarter AI model deployments, right? So, and in fact, you know, Anna just touched on this, but I want to, you know, make sure that it's clear that OctoML is putting that power into user's hands right now. So in partnership with AWS, we are launching this new PyTorch native profiler that allows you with a single, you know, one line, you know, code decorator allows you to see how your code runs on a variety of different hardware after accelerations. So it gives you very clear, you know, data on how you should think about your model deployments. And this ties back to choices of models. So like, if you have a set of choices that are equally good of models in terms of functionality and you want to understand after acceleration how are you going to deploy, how much they're going to cost or what are the options using a automated process of making a decision is really, really useful. And in fact, so I think these events can get early access to this by signing up for the Octopods, you know, this is exclusive group for insiders here, so you can go to OctoML.ai/pods to sign up. >> So that Octopod, is that a program? What is that, is that access to code? Is that a beta, what is that? Explain, take a minute and explain Octopod. >> I think the Octopod would be a group of people who is interested in experiencing this functionality. So it is the friends and users of OctoML that would be the Octopod. And then yes, after you sign up, we would provide you essentially the tool in code form for you to try out in your own. I mean, part of the benefit of this is that it happens in your own local environment and you're in control of everything kind of within the workflow that developers are already using to create and begin putting these models into their applications. So it would all be within your control. >> Got it. I think the big question I have for you is when do you, when does that one of your customers know they need to call you? What's their environment look like? What are they struggling with? What are the conversations they might be having on their side of the fence? If anyone's watching this, they're like, "Hey, you know what, I've got my team, we have a lot of data. Do we have our own language model or do I use someone else's?" There's a lot of this, I will say discovery going on around what to do, what path to take, what does that customer look like, if someone's listening, when do they know to call you guys, OctoML? >> Well, I mean the most obvious one is that you have a significant spend on AI/ML, come and talk to us, you know, putting AIML into production. So that's the clear one. In fact, just this morning I was talking to someone who is in life sciences space and is having, you know, 15 to $20 million a year cloud related to AI/ML deployment is a clear, it's a pretty clear match right there, right? So that's on the cost side. But I also want to emphasize something that Anna said earlier that, you know, the hardware and software complexity involved in putting model into production is really high. So we've been able to abstract that away, offering a clean automation flow enables one, to experiment early on, you know, how models would run and get them to production. And then two, once they are into production, gives you an automated flow to continuously updating your model and taking advantage of all this acceleration and ability to run the model on the right hardware. So anyways, let's say one then is cost, you know, you have significant cost and then two, you have an automation needs. And Anna please compliment that. >> Yeah, Anna you can please- >> Yeah, I think that's exactly right. Maybe the other time is when you are expecting a big scale up in serving your application, right? You're launching a new feature, you expect to get a lot of usage or, and you want to kind of anticipate maybe your CTO, your CIO, whoever pays your cloud bills is going to come after you, right? And so they want to know, you know, what's the return on putting this model essentially into my application stack? Am I going to, is the usage going to match what I'm paying for it? And then you can understand that. >> So you guys have a lot of the early adopters, they got big data teams, they're pushed in the production, they want to get a little QA, test the waters, understand, use your technology to figure it out. Is there any cases where people have gone into production, they have to pull it out? It's like the old lemon laws with your car, you buy a car and oh my god, it's not the way I wanted it. I mean, I can imagine the early people through the wall, so to speak, in the wave here are going to be bloody in the sense that they've gone in and tried stuff and get stuck with huge bills. Are you seeing that? Are people pulling stuff out of production and redeploying? Or I can imagine that if I had a bad deployment, I'd want to refactor that or actually replatform that. Do you see that too? >> Definitely after a sticker shock, yes, your customers will come and make sure that, you know, the sticker shock won't happen again. >> Yeah. >> But then there's another more thorough aspect here that I think we likely touched on, be worth elaborating a bit more is just how are you going to scale in a way that's feasible depending on the allocation that you get, right? So as we mentioned several times here, you know, model deployment is so hardware dependent and so complex that you tend to get a model for a hardware choice and then you want to scale that specific type of instance. But what if, when you want to scale because suddenly luckily got popular and, you know, you want to scale it up and then you don't have that instance anymore. So how do you live with whatever you have at that moment is something that we see customers needing as well. You know, so in fact, ideally what we want is customers to not think about what kind of specific instances they want. What they want is to know what their models need. Say, they know the SLA and then find a set of hybrid targets and instances that hit the SLA whenever they're also scaling, they're going to scale with more freedom, right? Instead of having to wait for AWS to give them more specific allocation for a specific instance. What if you could live with other types of hardware and scale up in a more free way, right? So that's another thing that we see customers, you know, like they need more freedom to be able to scale with whatever is available. >> Anna, you touched on this with the business model impact to that 6 million cost, if that goes out of control, there's a business model aspect and there's a technical operation aspect to the cost side too. You want to be mindful of riding the wave in a good way, but not getting over your skis. So that brings up the point around, you know, confidence, right? And teamwork. Because if you're in production, there's probably a team behind it. Talk about the team aspect of your customers. I mean, they're dedicated, they go put stuff into production, they're developers, there're data. What's in it for them? Are they getting better, are they in the beach, you know, reading the book. Are they, you know, are there easy street for them? What's the customer benefit to the teams? >> Yeah, absolutely. With just a few clicks of a button, you're in production, right? That's the dream. So yeah, I mean I think that, you know, we illustrated it before a little bit. I think the automated kind of benchmarking and optimization process, like when you think about the effort it takes to get that data by hand, which is what people are doing today, they just don't do it. So they're making decisions without the best information because it's, you know, there just isn't the bandwidth to get the information that they need to make the best decision and then know exactly how to deploy it. So I think it's actually bringing kind of a new insight and capability to these teams that they didn't have before. And then maybe another aspect on the team side is that it's making the hand-off of the models from the data science teams to the model deployment teams more seamless. So we have, you know, we have seen in the past that this kind of transition point is the place where there are a lot of hiccups, right? The data science team will give a model to the production team and it'll be too slow for the application or it'll be too expensive to run and it has to go back and be changed and kind of this loop. And so, you know, with the PyTorch profiler that Luis was talking about, and then also, you know, the other ways we do optimization that kind of prevents that hand-off problem from happening. >> Luis and Anna, you guys have a great company. Final couple minutes left. Talk about the company, the people there, what's the culture like, you know, if Intel has Moore's law, which is, you know, doubling the performance in few years, what's the culture like there? Is it, you know, more throughput, better pricing? Explain what's going on with the company and put a plug in. Luis, we'll start with you. >> Yeah, absolutely. I'm extremely proud of the team that we built here. You know, we have a people first culture, you know, very, very collaborative and folks, we all have a shared mission here of making AI more accessible and sustainable. We have a very diverse team in terms of backgrounds and life stories, you know, to do what we do here, we need a team that has expertise in software engineering, in machine learning, in computer architecture. Even though we don't build chips, we need to understand how they work, right? So, and then, you know, the fact that we have this, this very really, really varied set of backgrounds makes the environment, you know, it's say very exciting to learn more about, you know, assistance end-to-end. But also makes it for a very interesting, you know, work environment, right? So people have different backgrounds, different stories. Some of them went to grad school, others, you know, were in intelligence agencies and now are working here, you know. So we have a really interesting set of people and, you know, life is too short not to work with interesting humans. You know, that's something that I like to think about, you know. >> I'm sure your off-site meetings are a lot of fun, people talking about computer architectures, silicon advances, the next GPU, the big data models coming in. Anna, what's your take? What's the culture like? What's the company vibe and what are you guys looking to do? What's the customer success pattern? What's up? >> Yeah, absolutely. I mean, I, you know, second all of the great things that Luis just said about the team. I think one that I, an additional one that I'd really like to underscore is kind of this customer obsession, to use a term you all know well. And focus on the end users and really making the experiences that we're bringing to our user who are developers really, you know, useful and valuable for them. And so I think, you know, all of these tools that we're trying to put in the hands of users, the industry and the market is changing so rapidly that our products across the board, you know, all of the companies that, you know, are part of the showcase today, we're all evolving them so quickly and we can only do that kind of really hand in glove with our users. So that would be another thing I'd emphasize. >> I think the change dynamic, the power dynamics of this industry is just the beginning. I'm very bullish that this is going to be probably one of the biggest inflection points in history of the computer industry because of all the dynamics of the confluence of all the forces, which you mentioned some of them, I mean PC, you know, interoperability within internetworking and you got, you know, the web and then mobile. Now we have this, I mean, I wouldn't even put social media even in the close to this. Like, this is like, changes user experience, changes infrastructure. There's going to be massive accelerations in performance on the hardware side from AWS's of the world and cloud and you got the edge and more data. This is really what big data was going to look like. This is the beginning. Final question, what do you guys see going forward in the future? >> Well, it's undeniable that machine learning and AI models are becoming an integral part of an interesting application today, right? So, and the clear trends here are, you know, more and more competitional needs for these models because they're only getting more and more powerful. And then two, you know, seeing the complexity of the infrastructure where they run, you know, just considering the cloud, there's like a wide variety of choices there, right? So being able to live with that and making the most out of it in a way that does not require, you know, an impossible to find team is something that's pretty clear. So the need for automation, abstracting with the complexity is definitely here. And we are seeing this, you know, trends are that you also see models starting to move to the edge as well. So it's clear that we're seeing, we are going to live in a world where there's no large models living in the cloud. And then, you know, edge models that talk to these models in the cloud to form, you know, an end-to-end truly intelligent application. >> Anna? >> Yeah, I think, you know, our, Luis said it at the beginning. Our vision is to make AI sustainable and accessible. And I think as this technology just expands in every company and every team, that's going to happen kind of on its own. And we're here to help support that. And I think you can't do that without tools like those like OctoML. >> I think it's going to be an error of massive invention, creativity, a lot of the format heavy lifting is going to allow the talented people to automate their intellect. I mean, this is really kind of what we see going on. And Luis, thank you so much. Anna, thanks for coming on this segment. Thanks for coming on theCUBE and being part of the AWS Startup Showcase. I'm John Furrier, your host. Thanks for watching. (upbeat music)

(upbeat music) >> Welcome everyone to theCUBE's presentation of the AWS Startup Showcase AI/ML Top Startups Building Foundation Model Infrastructure. This is season three, episode one of our ongoing series covering exciting startups from the AWS ecosystem to talk about data and analytics. I'm your host, Lisa Martin and today we're excited to be joined by two guests from Astronomer. Steven Hillion joins us, it's Chief Data Officer and Jeff Fletcher, it's director of ML. They're here to talk about machine learning and data orchestration. Guys, thank you so much for joining us today. >> Thank you. >> It's great to be here. >> Before we get into machine learning let's give the audience an overview of Astronomer. Talk about what that is, Steven. Talk about what you mean by data orchestration. >> Yeah, let's start with Astronomer. We're the Airflow company basically. The commercial developer behind the open-source project, Apache Airflow. I don't know if you've heard of Airflow. It's sort of de-facto standard these days for orchestrating data pipelines, data engineering pipelines, and as we'll talk about later, machine learning pipelines. It's really is the de-facto standard. I think we're up to about 12 million downloads a month. That's actually as a open-source project. I think at this point it's more popular by some measures than Slack. Airflow was created by Airbnb some years ago to manage all of their data pipelines and manage all of their workflows and now it powers the data ecosystem for organizations as diverse as Electronic Arts, Conde Nast is one of our big customers, a big user of Airflow. And also not to mention the biggest banks on Wall Street use Airflow and Astronomer to power the flow of data throughout their organizations. >> Talk about that a little bit more, Steven, in terms of the business impact. You mentioned some great customer names there. What is the business impact or outcomes that a data orchestration strategy enables businesses to achieve? >> Yeah, I mean, at the heart of it is quite simply, scheduling and managing data pipelines. And so if you have some enormous retailer who's managing the flow of information throughout their organization they may literally have thousands or even tens of thousands of data pipelines that need to execute every day to do things as simple as delivering metrics for the executives to consume at the end of the day, to producing on a weekly basis new machine learning models that can be used to drive product recommendations. One of our customers, for example, is a British food delivery service. And you get those recommendations in your application that says, "Well, maybe you want to have samosas with your curry." That sort of thing is powered by machine learning models that they train on a regular basis to reflect changing conditions in the market. And those are produced through Airflow and through the Astronomer platform, which is essentially a managed platform for running airflow. So at its simplest it really is just scheduling and managing those workflows. But that's easier said than done of course. I mean if you have 10 thousands of those things then you need to make sure that they all run that they all have sufficient compute resources. If things fail, how do you track those down across those 10,000 workflows? How easy is it for an average data scientist or data engineer to contribute their code, their Python notebooks or their SQL code into a production environment? And then you've got reproducibility, governance, auditing, like managing data flows across an organization which we think of as orchestrating them is much more than just scheduling. It becomes really complicated pretty quickly. >> I imagine there's a fair amount of complexity there. Jeff, let's bring you into the conversation. Talk a little bit about Astronomer through your lens, data orchestration and how it applies to MLOps. >> So I come from a machine learning background and for me the interesting part is that machine learning requires the expansion into orchestration. A lot of the same things that you're using to go and develop and build pipelines in a standard data orchestration space applies equally well in a machine learning orchestration space. What you're doing is you're moving data between different locations, between different tools, and then tasking different types of tools to act on that data. So extending it made logical sense from a implementation perspective. And a lot of my focus at Astronomer is really to explain how Airflow can be used well in a machine learning context. It is being used well, it is being used a lot by the customers that we have and also by users of the open source version. But it's really being able to explain to people why it's a natural extension for it and how well it fits into that. And a lot of it is also extending some of the infrastructure capabilities that Astronomer provides to those customers for them to be able to run some of the more platform specific requirements that come with doing machine learning pipelines. >> Let's get into some of the things that make Astronomer unique. Jeff, sticking with you, when you're in customer conversations, what are some of the key differentiators that you articulate to customers? >> So a lot of it is that we are not specific to one cloud provider. So we have the ability to operate across all of the big cloud providers. I know, I'm certain we have the best developers that understand how best practices implementations for data orchestration works. So we spend a lot of time talking to not just the business outcomes and the business users of the product, but also also for the technical people, how to help them better implement things that they may have come across on a Stack Overflow article or not necessarily just grown with how the product has migrated. So it's the ability to run it wherever you need to run it and also our ability to help you, the customer, better implement and understand those workflows that I think are two of the primary differentiators that we have. >> Lisa: Got it. >> I'll add another one if you don't mind. >> You can go ahead, Steven. >> Is lineage and dependencies between workflows. One thing we've done is to augment core Airflow with Lineage services. So using the Open Lineage framework, another open source framework for tracking datasets as they move from one workflow to another one, team to another, one data source to another is a really key component of what we do and we bundle that within the service so that as a developer or as a production engineer, you really don't have to worry about lineage, it just happens. Jeff, may show us some of this later that you can actually see as data flows from source through to a data warehouse out through a Python notebook to produce a predictive model or a dashboard. Can you see how those data products relate to each other? And when something goes wrong, figure out what upstream maybe caused the problem, or if you're about to change something, figure out what the impact is going to be on the rest of the organization. So Lineage is a big deal for us. >> Got it. >> And just to add on to that, the other thing to think about is that traditional Airflow is actually a complicated implementation. It required quite a lot of time spent understanding or was almost a bespoke language that you needed to be able to develop in two write these DAGs, which is like fundamental pipelines. So part of what we are focusing on is tooling that makes it more accessible to say a data analyst or a data scientist who doesn't have or really needs to gain the necessary background in how the semantics of Airflow DAGs works to still be able to get the benefit of what Airflow can do. So there is new features and capabilities built into the astronomer cloud platform that effectively obfuscates and removes the need to understand some of the deep work that goes on. But you can still do it, you still have that capability, but we are expanding it to be able to have orchestrated and repeatable processes accessible to more teams within the business. >> In terms of accessibility to more teams in the business. You talked about data scientists, data analysts, developers. Steven, I want to talk to you, as the chief data officer, are you having more and more conversations with that role and how is it emerging and evolving within your customer base? >> Hmm. That's a good question, and it is evolving because I think if you look historically at the way that Airflow has been used it's often from the ground up. You have individual data engineers or maybe single data engineering teams who adopt Airflow 'cause it's very popular. Lots of people know how to use it and they bring it into an organization and say, "Hey, let's use this to run our data pipelines." But then increasingly as you turn from pure workflow management and job scheduling to the larger topic of orchestration you realize it gets pretty complicated, you want to have coordination across teams, and you want to have standardization for the way that you manage your data pipelines. And so having a managed service for Airflow that exists in the cloud is easy to spin up as you expand usage across the organization. And thinking long term about that in the context of orchestration that's where I think the chief data officer or the head of analytics tends to get involved because they really want to think of this as a strategic investment that they're making. Not just per team individual Airflow deployments, but a network of data orchestrators. >> That network is key. Every company these days has to be a data company. We talk about companies being data driven. It's a common word, but it's true. It's whether it is a grocer or a bank or a hospital, they've got to be data companies. So talk to me a little bit about Astronomer's business model. How is this available? How do customers get their hands on it? >> Jeff, go ahead. >> Yeah, yeah. So we have a managed cloud service and we have two modes of operation. One, you can bring your own cloud infrastructure. So you can say here is an account in say, AWS or Azure and we can go and deploy the necessary infrastructure into that, or alternatively we can host everything for you. So it becomes a full SaaS offering. But we then provide a platform that connects at the backend to your internal IDP process. So however you are authenticating users to make sure that the correct people are accessing the services that they need with role-based access control. From there we are deploying through Kubernetes, the different services and capabilities into either your cloud account or into an account that we host. And from there Airflow does what Airflow does, which is its ability to then reach to different data systems and data platforms and to then run the orchestration. We make sure we do it securely, we have all the necessary compliance certifications required for GDPR in Europe and HIPAA based out of the US, and a whole bunch host of others. So it is a secure platform that can run in a place that you need it to run, but it is a managed Airflow that includes a lot of the extra capabilities like the cloud developer environment and the open lineage services to enhance the overall airflow experience. >> Enhance the overall experience. So Steven, going back to you, if I'm a Conde Nast or another organization, what are some of the key business outcomes that I can expect? As one of the things I think we've learned during the pandemic is access to realtime data is no longer a nice to have for organizations. It's really an imperative. It's that demanding consumer that wants to have that personalized, customized, instant access to a product or a service. So if I'm a Conde Nast or I'm one of your customers, what can I expect my business to be able to achieve as a result of data orchestration? >> Yeah, I think in a nutshell it's about providing a reliable, scalable, and easy to use service for developing and running data workflows. And talking of demanding customers, I mean, I'm actually a customer myself, as you mentioned, I'm the head of data for Astronomer. You won't be surprised to hear that we actually use Astronomer and Airflow to run all of our data pipelines. And so I can actually talk about my experience. When I started I was of course familiar with Airflow, but it always seemed a little bit unapproachable to me if I was introducing that to a new team of data scientists. They don't necessarily want to have to think about learning something new. But I think because of the layers that Astronomer has provided with our Astro service around Airflow it was pretty easy for me to get up and running. Of course I've got an incentive for doing that. I work for the Airflow company, but we went from about, at the beginning of last year, about 500 data tasks that we were running on a daily basis to about 15,000 every day. We run something like a million data operations every month within my team. And so as one outcome, just the ability to spin up new production workflows essentially in a single day you go from an idea in the morning to a new dashboard or a new model in the afternoon, that's really the business outcome is just removing that friction to operationalizing your machine learning and data workflows. >> And I imagine too, oh, go ahead, Jeff. >> Yeah, I think to add to that, one of the things that becomes part of the business cycle is a repeatable capabilities for things like reporting, for things like new machine learning models. And the impediment that has existed is that it's difficult to take that from a team that's an analyst team who then provide that or a data science team that then provide that to the data engineering team who have to work the workflow all the way through. What we're trying to unlock is the ability for those teams to directly get access to scheduling and orchestrating capabilities so that a business analyst can have a new report for C-suite execs that needs to be done once a week, but the time to repeatability for that report is much shorter. So it is then immediately in the hands of the person that needs to see it. It doesn't have to go into a long list of to-dos for a data engineering team that's already overworked that they eventually get it to it in a month's time. So that is also a part of it is that the realizing, orchestration I think is fairly well and a lot of people get the benefit of being able to orchestrate things within a business, but it's having more people be able to do it and shorten the time that that repeatability is there is one of the main benefits from good managed orchestration. >> So a lot of workforce productivity improvements in what you're doing to simplify things, giving more people access to data to be able to make those faster decisions, which ultimately helps the end user on the other end to get that product or the service that they're expecting like that. Jeff, I understand you have a demo that you can share so we can kind of dig into this. >> Yeah, let me take you through a quick look of how the whole thing works. So our starting point is our cloud infrastructure. This is the login. You go to the portal. You can see there's a a bunch of workspaces that are available. Workspaces are like individual places for people to operate in. I'm not going to delve into all the deep technical details here, but starting point for a lot of our data science customers is we have what we call our Cloud IDE, which is a web-based development environment for writing and building out DAGs without actually having to know how the underpinnings of Airflow work. This is an internal one, something that we use. You have a notebook-like interface that lets you write python code and SQL code and a bunch of specific bespoke type of blocks if you want. They all get pulled together and create a workflow. So this is a workflow, which gets compiled to something that looks like a complicated set of Python code, which is the DAG. I then have a CICD process pipeline where I commit this through to my GitHub repo. So this comes to a repo here, which is where these DAGs that I created in the previous step exist. I can then go and say, all right, I want to see how those particular DAGs have been running. We then get to the actual Airflow part. So this is the managed Airflow component. So we add the ability for teams to fairly easily bring up an Airflow instance and write code inside our notebook-like environment to get it into that instance. So you can see it's been running. That same process that we built here that graph ends up here inside this, but you don't need to know how the fundamentals of Airflow work in order to get this going. Then we can run one of these, it runs in the background and we can manage how it goes. And from there, every time this runs, it's emitting to a process underneath, which is the open lineage service, which is the lineage integration that allows me to come in here and have a look and see this was that actual, that same graph that we built, but now it's the historic version. So I know where things started, where things are going, and how it ran. And then I can also do a comparison. So if I want to see how this particular run worked compared to one historically, I can grab one from a previous date and it will show me the comparison between the two. So that combination of managed Airflow, getting Airflow up and running very quickly, but the Cloud IDE that lets you write code and know how to get something into a repeatable format get that into Airflow and have that attached to the lineage process adds what is a complete end-to-end orchestration process for any business looking to get the benefit from orchestration. >> Outstanding. Thank you so much Jeff for digging into that. So one of my last questions, Steven is for you. This is exciting. There's a lot that you guys are enabling organizations to achieve here to really become data-driven companies. So where can folks go to get their hands on this? >> Yeah, just go to astronomer.io and we have plenty of resources. If you're new to Airflow, you can read our documentation, our guides to getting started. We have a CLI that you can download that is really I think the easiest way to get started with Airflow. But you can actually sign up for a trial. You can sign up for a guided trial where our teams, we have a team of experts, really the world experts on getting Airflow up and running. And they'll take you through that trial and allow you to actually kick the tires and see how this works with your data. And I think you'll see pretty quickly that it's very easy to get started with Airflow, whether you're doing that from the command line or doing that in our cloud service. And all of that is available on our website >> astronomer.io. Jeff, last question for you. What are you excited about? There's so much going on here. What are some of the things, maybe you can give us a sneak peek coming down the road here that prospects and existing customers should be excited about? >> I think a lot of the development around the data awareness components, so one of the things that's traditionally been complicated with orchestration is you leave your data in the place that you're operating on and we're starting to have more data processing capability being built into Airflow. And from a Astronomer perspective, we are adding more capabilities around working with larger datasets, doing bigger data manipulation with inside the Airflow process itself. And that lends itself to better machine learning implementation. So as we start to grow and as we start to get better in the machine learning context, well, in the data awareness context, it unlocks a lot more capability to do and implement proper machine learning pipelines. >> Awesome guys. Exciting stuff. Thank you so much for talking to me about Astronomer, machine learning, data orchestration, and really the value in it for your customers. Steve and Jeff, we appreciate your time. >> Thank you. >> My pleasure, thanks. >> And we thank you for watching. This is season three, episode one of our ongoing series covering exciting startups from the AWS ecosystem. I'm your host, Lisa Martin. You're watching theCUBE, the leader in live tech coverage. (upbeat music)

(upbeat music) >> Hello everyone. Welcome to theCube's presentation of the "AWS Startup Showcase." The topic this episode is AI and machine learning, top startups building foundational model infrastructure. This is season three, episode one of the ongoing series covering exciting startups from the AWS ecosystem. And this time we're talking about AI and machine learning. I'm your host, John Furrier. I'm excited I'm joined today by Robert Nishihara, who's the co-founder and CEO of a hot startup called Anyscale. He's here to talk about Ray, the open source project, Anyscale's infrastructure for foundation as well. Robert, thank you for joining us today. >> Yeah, thanks so much as well. >> I've been following your company since the founding pre pandemic and you guys really had a great vision scaled up and in a perfect position for this big wave that we all see with ChatGPT and OpenAI that's gone mainstream. Finally, AI has broken out through the ropes and now gone mainstream, so I think you guys are really well positioned. I'm looking forward to to talking with you today. But before we get into it, introduce the core mission for Anyscale. Why do you guys exist? What is the North Star for Anyscale? >> Yeah, like you mentioned, there's a tremendous amount of excitement about AI right now. You know, I think a lot of us believe that AI can transform just every different industry. So one of the things that was clear to us when we started this company was that the amount of compute needed to do AI was just exploding. Like to actually succeed with AI, companies like OpenAI or Google or you know, these companies getting a lot of value from AI, were not just running these machine learning models on their laptops or on a single machine. They were scaling these applications across hundreds or thousands or more machines and GPUs and other resources in the Cloud. And so to actually succeed with AI, and this has been one of the biggest trends in computing, maybe the biggest trend in computing in, you know, in recent history, the amount of compute has been exploding. And so to actually succeed with that AI, to actually build these scalable applications and scale the AI applications, there's a tremendous software engineering lift to build the infrastructure to actually run these scalable applications. And that's very hard to do. So one of the reasons many AI projects and initiatives fail is that, or don't make it to production, is the need for this scale, the infrastructure lift, to actually make it happen. So our goal here with Anyscale and Ray, is to make that easy, is to make scalable computing easy. So that as a developer or as a business, if you want to do AI, if you want to get value out of AI, all you need to know is how to program on your laptop. Like, all you need to know is how to program in Python. And if you can do that, then you're good to go. Then you can do what companies like OpenAI or Google do and get value out of machine learning. >> That programming example of how easy it is with Python reminds me of the early days of Cloud, when infrastructure as code was talked about was, it was just code the infrastructure programmable. That's super important. That's what AI people wanted, first program AI. That's the new trend. And I want to understand, if you don't mind explaining, the relationship that Anyscale has to these foundational models and particular the large language models, also called LLMs, was seen with like OpenAI and ChatGPT. Before you get into the relationship that you have with them, can you explain why the hype around foundational models? Why are people going crazy over foundational models? What is it and why is it so important? >> Yeah, so foundational models and foundation models are incredibly important because they enable businesses and developers to get value out of machine learning, to use machine learning off the shelf with these large models that have been trained on tons of data and that are useful out of the box. And then, of course, you know, as a business or as a developer, you can take those foundational models and repurpose them or fine tune them or adapt them to your specific use case and what you want to achieve. But it's much easier to do that than to train them from scratch. And I think there are three, for people to actually use foundation models, there are three main types of workloads or problems that need to be solved. One is training these foundation models in the first place, like actually creating them. The second is fine tuning them and adapting them to your use case. And the third is serving them and actually deploying them. Okay, so Ray and Anyscale are used for all of these three different workloads. Companies like OpenAI or Cohere that train large language models. Or open source versions like GPTJ are done on top of Ray. There are many startups and other businesses that fine tune, that, you know, don't want to train the large underlying foundation models, but that do want to fine tune them, do want to adapt them to their purposes, and build products around them and serve them, those are also using Ray and Anyscale for that fine tuning and that serving. And so the reason that Ray and Anyscale are important here is that, you know, building and using foundation models requires a huge scale. It requires a lot of data. It requires a lot of compute, GPUs, TPUs, other resources. And to actually take advantage of that and actually build these scalable applications, there's a lot of infrastructure that needs to happen under the hood. And so you can either use Ray and Anyscale to take care of that and manage the infrastructure and solve those infrastructure problems. Or you can build the infrastructure and manage the infrastructure yourself, which you can do, but it's going to slow your team down. It's going to, you know, many of the businesses we work with simply don't want to be in the business of managing infrastructure and building infrastructure. They want to focus on product development and move faster. >> I know you got a keynote presentation we're going to go to in a second, but I think you hit on something I think is the real tipping point, doing it yourself, hard to do. These are things where opportunities are and the Cloud did that with data centers. Turned a data center and made it an API. The heavy lifting went away and went to the Cloud so people could be more creative and build their product. In this case, build their creativity. Is that kind of what's the big deal? Is that kind of a big deal happening that you guys are taking the learnings and making that available so people don't have to do that? >> That's exactly right. So today, if you want to succeed with AI, if you want to use AI in your business, infrastructure work is on the critical path for doing that. To do AI, you have to build infrastructure. You have to figure out how to scale your applications. That's going to change. We're going to get to the point, and you know, with Ray and Anyscale, we're going to remove the infrastructure from the critical path so that as a developer or as a business, all you need to focus on is your application logic, what you want the the program to do, what you want your application to do, how you want the AI to actually interface with the rest of your product. Now the way that will happen is that Ray and Anyscale will still, the infrastructure work will still happen. It'll just be under the hood and taken care of by Ray in Anyscale. And so I think something like this is really necessary for AI to reach its potential, for AI to have the impact and the reach that we think it will, you have to make it easier to do. >> And just for clarification to point out, if you don't mind explaining the relationship of Ray and Anyscale real quick just before we get into the presentation. >> So Ray is an open source project. We created it. We were at Berkeley doing machine learning. We started Ray so that, in order to provide an easy, a simple open source tool for building and running scalable applications. And Anyscale is the managed version of Ray, basically we will run Ray for you in the Cloud, provide a lot of tools around the developer experience and managing the infrastructure and providing more performance and superior infrastructure. >> Awesome. I know you got a presentation on Ray and Anyscale and you guys are positioning as the infrastructure for foundational models. So I'll let you take it away and then when you're done presenting, we'll come back, I'll probably grill you with a few questions and then we'll close it out so take it away. >> Robert: Sounds great. So I'll say a little bit about how companies are using Ray and Anyscale for foundation models. The first thing I want to mention is just why we're doing this in the first place. And the underlying observation, the underlying trend here, and this is a plot from OpenAI, is that the amount of compute needed to do machine learning has been exploding. It's been growing at something like 35 times every 18 months. This is absolutely enormous. And other people have written papers measuring this trend and you get different numbers. But the point is, no matter how you slice and dice it, it' a astronomical rate. Now if you compare that to something we're all familiar with, like Moore's Law, which says that, you know, the processor performance doubles every roughly 18 months, you can see that there's just a tremendous gap between the needs, the compute needs of machine learning applications, and what you can do with a single chip, right. So even if Moore's Law were continuing strong and you know, doing what it used to be doing, even if that were the case, there would still be a tremendous gap between what you can do with the chip and what you need in order to do machine learning. And so given this graph, what we've seen, and what has been clear to us since we started this company, is that doing AI requires scaling. There's no way around it. It's not a nice to have, it's really a requirement. And so that led us to start Ray, which is the open source project that we started to make it easy to build these scalable Python applications and scalable machine learning applications. And since we started the project, it's been adopted by a tremendous number of companies. Companies like OpenAI, which use Ray to train their large models like ChatGPT, companies like Uber, which run all of their deep learning and classical machine learning on top of Ray, companies like Shopify or Spotify or Instacart or Lyft or Netflix, ByteDance, which use Ray for their machine learning infrastructure. Companies like Ant Group, which makes Alipay, you know, they use Ray across the board for fraud detection, for online learning, for detecting money laundering, you know, for graph processing, stream processing. Companies like Amazon, you know, run Ray at a tremendous scale and just petabytes of data every single day. And so the project has seen just enormous adoption since, over the past few years. And one of the most exciting use cases is really providing the infrastructure for building training, fine tuning, and serving foundation models. So I'll say a little bit about, you know, here are some examples of companies using Ray for foundation models. Cohere trains large language models. OpenAI also trains large language models. You can think about the workloads required there are things like supervised pre-training, also reinforcement learning from human feedback. So this is not only the regular supervised learning, but actually more complex reinforcement learning workloads that take human input about what response to a particular question, you know is better than a certain other response. And incorporating that into the learning. There's open source versions as well, like GPTJ also built on top of Ray as well as projects like Alpa coming out of UC Berkeley. So these are some of the examples of exciting projects in organizations, training and creating these large language models and serving them using Ray. Okay, so what actually is Ray? Well, there are two layers to Ray. At the lowest level, there's the core Ray system. This is essentially low level primitives for building scalable Python applications. Things like taking a Python function or a Python class and executing them in the cluster setting. So Ray core is extremely flexible and you can build arbitrary scalable applications on top of Ray. So on top of Ray, on top of the core system, what really gives Ray a lot of its power is this ecosystem of scalable libraries. So on top of the core system you have libraries, scalable libraries for ingesting and pre-processing data, for training your models, for fine tuning those models, for hyper parameter tuning, for doing batch processing and batch inference, for doing model serving and deployment, right. And a lot of the Ray users, the reason they like Ray is that they want to run multiple workloads. They want to train and serve their models, right. They want to load their data and feed that into training. And Ray provides common infrastructure for all of these different workloads. So this is a little overview of what Ray, the different components of Ray. So why do people choose to go with Ray? I think there are three main reasons. The first is the unified nature. The fact that it is common infrastructure for scaling arbitrary workloads, from data ingest to pre-processing to training to inference and serving, right. This also includes the fact that it's future proof. AI is incredibly fast moving. And so many people, many companies that have built their own machine learning infrastructure and standardized on particular workflows for doing machine learning have found that their workflows are too rigid to enable new capabilities. If they want to do reinforcement learning, if they want to use graph neural networks, they don't have a way of doing that with their standard tooling. And so Ray, being future proof and being flexible and general gives them that ability. Another reason people choose Ray in Anyscale is the scalability. This is really our bread and butter. This is the reason, the whole point of Ray, you know, making it easy to go from your laptop to running on thousands of GPUs, making it easy to scale your development workloads and run them in production, making it easy to scale, you know, training to scale data ingest, pre-processing and so on. So scalability and performance, you know, are critical for doing machine learning and that is something that Ray provides out of the box. And lastly, Ray is an open ecosystem. You can run it anywhere. You can run it on any Cloud provider. Google, you know, Google Cloud, AWS, Asure. You can run it on your Kubernetes cluster. You can run it on your laptop. It's extremely portable. And not only that, it's framework agnostic. You can use Ray to scale arbitrary Python workloads. You can use it to scale and it integrates with libraries like TensorFlow or PyTorch or JAX or XG Boost or Hugging Face or PyTorch Lightning, right, or Scikit-learn or just your own arbitrary Python code. It's open source. And in addition to integrating with the rest of the machine learning ecosystem and these machine learning frameworks, you can use Ray along with all of the other tooling in the machine learning ecosystem. That's things like weights and biases or ML flow, right. Or you know, different data platforms like Databricks, you know, Delta Lake or Snowflake or tools for model monitoring for feature stores, all of these integrate with Ray. And that's, you know, Ray provides that kind of flexibility so that you can integrate it into the rest of your workflow. And then Anyscale is the scalable compute platform that's built on top, you know, that provides Ray. So Anyscale is a managed Ray service that runs in the Cloud. And what Anyscale does is it offers the best way to run Ray. And if you think about what you get with Anyscale, there are fundamentally two things. One is about moving faster, accelerating the time to market. And you get that by having the managed service so that as a developer you don't have to worry about managing infrastructure, you don't have to worry about configuring infrastructure. You also, it provides, you know, optimized developer workflows. Things like easily moving from development to production, things like having the observability tooling, the debug ability to actually easily diagnose what's going wrong in a distributed application. So things like the dashboards and the other other kinds of tooling for collaboration, for monitoring and so on. And then on top of that, so that's the first bucket, developer productivity, moving faster, faster experimentation and iteration. The second reason that people choose Anyscale is superior infrastructure. So this is things like, you know, cost deficiency, being able to easily take advantage of spot instances, being able to get higher GPU utilization, things like faster cluster startup times and auto scaling. Things like just overall better performance and faster scheduling. And so these are the kinds of things that Anyscale provides on top of Ray. It's the managed infrastructure. It's fast, it's like the developer productivity and velocity as well as performance. So this is what I wanted to share about Ray in Anyscale. >> John: Awesome. >> Provide that context. But John, I'm curious what you think. >> I love it. I love the, so first of all, it's a platform because that's the platform architecture right there. So just to clarify, this is an Anyscale platform, not- >> That's right. >> Tools. So you got tools in the platform. Okay, that's key. Love that managed service. Just curious, you mentioned Python multiple times, is that because of PyTorch and TensorFlow or Python's the most friendly with machine learning or it's because it's very common amongst all developers? >> That's a great question. Python is the language that people are using to do machine learning. So it's the natural starting point. Now, of course, Ray is actually designed in a language agnostic way and there are companies out there that use Ray to build scalable Java applications. But for the most part right now we're focused on Python and being the best way to build these scalable Python and machine learning applications. But, of course, down the road there always is that potential. >> So if you're slinging Python code out there and you're watching that, you're watching this video, get on Anyscale bus quickly. Also, I just, while you were giving the presentation, I couldn't help, since you mentioned OpenAI, which by the way, congratulations 'cause they've had great scale, I've noticed in their rapid growth 'cause they were the fastest company to the number of users than anyone in the history of the computer industry, so major successor, OpenAI and ChatGPT, huge fan. I'm not a skeptic at all. I think it's just the beginning, so congratulations. But I actually typed into ChatGPT, what are the top three benefits of Anyscale and came up with scalability, flexibility, and ease of use. Obviously, scalability is what you guys are called. >> That's pretty good. >> So that's what they came up with. So they nailed it. Did you have an inside prompt training, buy it there? Only kidding. (Robert laughs) >> Yeah, we hard coded that one. >> But that's the kind of thing that came up really, really quickly if I asked it to write a sales document, it probably will, but this is the future interface. This is why people are getting excited about the foundational models and the large language models because it's allowing the interface with the user, the consumer, to be more human, more natural. And this is clearly will be in every application in the future. >> Absolutely. This is how people are going to interface with software, how they're going to interface with products in the future. It's not just something, you know, not just a chat bot that you talk to. This is going to be how you get things done, right. How you use your web browser or how you use, you know, how you use Photoshop or how you use other products. Like you're not going to spend hours learning all the APIs and how to use them. You're going to talk to it and tell it what you want it to do. And of course, you know, if it doesn't understand it, it's going to ask clarifying questions. You're going to have a conversation and then it'll figure it out. >> This is going to be one of those things, we're going to look back at this time Robert and saying, "Yeah, from that company, that was the beginning of that wave." And just like AWS and Cloud Computing, the folks who got in early really were in position when say the pandemic came. So getting in early is a good thing and that's what everyone's talking about is getting in early and playing around, maybe replatforming or even picking one or few apps to refactor with some staff and managed services. So people are definitely jumping in. So I have to ask you the ROI cost question. You mentioned some of those, Moore's Law versus what's going on in the industry. When you look at that kind of scale, the first thing that jumps out at people is, "Okay, I love it. Let's go play around." But what's it going to cost me? Am I going to be tied to certain GPUs? What's the landscape look like from an operational standpoint, from the customer? Are they locked in and the benefit was flexibility, are you flexible to handle any Cloud? What is the customers, what are they looking at? Basically, that's my question. What's the customer looking at? >> Cost is super important here and many of the companies, I mean, companies are spending a huge amount on their Cloud computing, on AWS, and on doing AI, right. And I think a lot of the advantage of Anyscale, what we can provide here is not only better performance, but cost efficiency. Because if we can run something faster and more efficiently, it can also use less resources and you can lower your Cloud spending, right. We've seen companies go from, you know, 20% GPU utilization with their current setup and the current tools they're using to running on Anyscale and getting more like 95, you know, 100% GPU utilization. That's something like a five x improvement right there. So depending on the kind of application you're running, you know, it's a significant cost savings. We've seen companies that have, you know, processing petabytes of data every single day with Ray going from, you know, getting order of magnitude cost savings by switching from what they were previously doing to running their application on Ray. And when you have applications that are spending, you know, potentially $100 million a year and getting a 10 X cost savings is just absolutely enormous. So these are some of the kinds of- >> Data infrastructure is super important. Again, if the customer, if you're a prospect to this and thinking about going in here, just like the Cloud, you got infrastructure, you got the platform, you got SaaS, same kind of thing's going to go on in AI. So I want to get into that, you know, ROI discussion and some of the impact with your customers that are leveraging the platform. But first I hear you got a demo. >> Robert: Yeah, so let me show you, let me give you a quick run through here. So what I have open here is the Anyscale UI. I've started a little Anyscale Workspace. So Workspaces are the Anyscale concept for interactive developments, right. So here, imagine I'm just, you want to have a familiar experience like you're developing on your laptop. And here I have a terminal. It's not on my laptop. It's actually in the cloud running on Anyscale. And I'm just going to kick this off. This is going to train a large language model, so OPT. And it's doing this on 32 GPUs. We've got a cluster here with a bunch of CPU cores, bunch of memory. And as that's running, and by the way, if I wanted to run this on instead of 32 GPUs, 64, 128, this is just a one line change when I launch the Workspace. And what I can do is I can pull up VS code, right. Remember this is the interactive development experience. I can look at the actual code. Here it's using Ray train to train the torch model. We've got the training loop and we're saying that each worker gets access to one GPU and four CPU cores. And, of course, as I make the model larger, this is using deep speed, as I make the model larger, I could increase the number of GPUs that each worker gets access to, right. And how that is distributed across the cluster. And if I wanted to run on CPUs instead of GPUs or a different, you know, accelerator type, again, this is just a one line change. And here we're using Ray train to train the models, just taking my vanilla PyTorch model using Hugging Face and then scaling that across a bunch of GPUs. And, of course, if I want to look at the dashboard, I can go to the Ray dashboard. There are a bunch of different visualizations I can look at. I can look at the GPU utilization. I can look at, you know, the CPU utilization here where I think we're currently loading the model and running that actual application to start the training. And some of the things that are really convenient here about Anyscale, both I can get that interactive development experience with VS code. You know, I can look at the dashboards. I can monitor what's going on. It feels, I have a terminal, it feels like my laptop, but it's actually running on a large cluster. And I can, with however many GPUs or other resources that I want. And so it's really trying to combine the best of having the familiar experience of programming on your laptop, but with the benefits, you know, being able to take advantage of all the resources in the Cloud to scale. And it's like when, you know, you're talking about cost efficiency. One of the biggest reasons that people waste money, one of the silly reasons for wasting money is just forgetting to turn off your GPUs. And what you can do here is, of course, things will auto terminate if they're idle. But imagine you go to sleep, I have this big cluster. You can turn it off, shut off the cluster, come back tomorrow, restart the Workspace, and you know, your big cluster is back up and all of your code changes are still there. All of your local file edits. It's like you just closed your laptop and came back and opened it up again. And so this is the kind of experience we want to provide for our users. So that's what I wanted to share with you. >> Well, I think that whole, couple of things, lines of code change, single line of code change, that's game changing. And then the cost thing, I mean human error is a big deal. People pass out at their computer. They've been coding all night or they just forget about it. I mean, and then it's just like leaving the lights on or your water running in your house. It's just, at the scale that it is, the numbers will add up. That's a huge deal. So I think, you know, compute back in the old days, there's no compute. Okay, it's just compute sitting there idle. But you know, data cranking the models is doing, that's a big point. >> Another thing I want to add there about cost efficiency is that we make it really easy to use, if you're running on Anyscale, to use spot instances and these preemptable instances that can just be significantly cheaper than the on-demand instances. And so when we see our customers go from what they're doing before to using Anyscale and they go from not using these spot instances 'cause they don't have the infrastructure around it, the fault tolerance to handle the preemption and things like that, to being able to just check a box and use spot instances and save a bunch of money. >> You know, this was my whole, my feature article at Reinvent last year when I met with Adam Selipsky, this next gen Cloud is here. I mean, it's not auto scale, it's infrastructure scale. It's agility. It's flexibility. I think this is where the world needs to go. Almost what DevOps did for Cloud and what you were showing me that demo had this whole SRE vibe. And remember Google had site reliability engines to manage all those servers. This is kind of like an SRE vibe for data at scale. I mean, a similar kind of order of magnitude. I mean, I might be a little bit off base there, but how would you explain it? >> It's a nice analogy. I mean, what we are trying to do here is get to the point where developers don't think about infrastructure. Where developers only think about their application logic. And where businesses can do AI, can succeed with AI, and build these scalable applications, but they don't have to build, you know, an infrastructure team. They don't have to develop that expertise. They don't have to invest years in building their internal machine learning infrastructure. They can just focus on the Python code, on their application logic, and run the stuff out of the box. >> Awesome. Well, I appreciate the time. Before we wrap up here, give a plug for the company. I know you got a couple websites. Again, go, Ray's got its own website. You got Anyscale. You got an event coming up. Give a plug for the company looking to hire. Put a plug in for the company. >> Yeah, absolutely. Thank you. So first of all, you know, we think AI is really going to transform every industry and the opportunity is there, right. We can be the infrastructure that enables all of that to happen, that makes it easy for companies to succeed with AI, and get value out of AI. Now we have, if you're interested in learning more about Ray, Ray has been emerging as the standard way to build scalable applications. Our adoption has been exploding. I mentioned companies like OpenAI using Ray to train their models. But really across the board companies like Netflix and Cruise and Instacart and Lyft and Uber, you know, just among tech companies. It's across every industry. You know, gaming companies, agriculture, you know, farming, robotics, drug discovery, you know, FinTech, we see it across the board. And all of these companies can get value out of AI, can really use AI to improve their businesses. So if you're interested in learning more about Ray and Anyscale, we have our Ray Summit coming up in September. This is going to highlight a lot of the most impressive use cases and stories across the industry. And if your business, if you want to use LLMs, you want to train these LLMs, these large language models, you want to fine tune them with your data, you want to deploy them, serve them, and build applications and products around them, give us a call, talk to us. You know, we can really take the infrastructure piece, you know, off the critical path and make that easy for you. So that's what I would say. And, you know, like you mentioned, we're hiring across the board, you know, engineering, product, go-to-market, and it's an exciting time. >> Robert Nishihara, co-founder and CEO of Anyscale, congratulations on a great company you've built and continuing to iterate on and you got growth ahead of you, you got a tailwind. I mean, the AI wave is here. I think OpenAI and ChatGPT, a customer of yours, have really opened up the mainstream visibility into this new generation of applications, user interface, roll of data, large scale, how to make that programmable so we're going to need that infrastructure. So thanks for coming on this season three, episode one of the ongoing series of the hot startups. In this case, this episode is the top startups building foundational model infrastructure for AI and ML. I'm John Furrier, your host. Thanks for watching. (upbeat music)

(light airy music) >> Hello, everyone, welcome to theCUBE's presentation of the AWS Startup Showcase, AI and machine learning. "Top Startups Building Generative AI on AWS." This is season three, episode one of the ongoing series covering the exciting startups from the AWS ecosystem, talking about AI machine learning. We have three great guests Bratin Saha, VP, Vice President of Machine Learning and AI Services at Amazon Web Services. Tom Mason, the CTO of Stability AI, and Aidan Gomez, CEO and co-founder of Cohere. Two practitioners doing startups and AWS. Gentlemen, thank you for opening up this session, this episode. Thanks for coming on. >> Thank you. >> Thank you. >> Thank you. >> So the topic is hype versus reality. So I think we're all on the reality is great, hype is great, but the reality's here. I want to get into it. Generative AI's got all the momentum, it's going mainstream, it's kind of come out of the behind the ropes, it's now mainstream. We saw the success of ChatGPT, opens up everyone's eyes, but there's so much more going on. Let's jump in and get your early perspectives on what should people be talking about right now? What are you guys working on? We'll start with AWS. What's the big focus right now for you guys as you come into this market that's highly active, highly hyped up, but people see value right out of the gate? >> You know, we have been working on generative AI for some time. In fact, last year we released Code Whisperer, which is about using generative AI for software development and a number of customers are using it and getting real value out of it. So generative AI is now something that's mainstream that can be used by enterprise users. And we have also been partnering with a number of other companies. So, you know, stability.ai, we've been partnering with them a lot. We want to be partnering with other companies as well. In seeing how we do three things, you know, first is providing the most efficient infrastructure for generative AI. And that is where, you know, things like Trainium, things like Inferentia, things like SageMaker come in. And then next is the set of models and then the third is the kind of applications like Code Whisperer and so on. So, you know, it's early days yet, but clearly there's a lot of amazing capabilities that will come out and something that, you know, our customers are starting to pay a lot of attention to. >> Tom, talk about your company and what your focus is and why the Amazon Web Services relationship's important for you? >> So yeah, we're primarily committed to making incredible open source foundation models and obviously stable effusions been our kind of first big model there, which we trained all on AWS. We've been working with them over the last year and a half to develop, obviously a big cluster, and bring all that compute to training these models at scale, which has been a really successful partnership. And we're excited to take it further this year as we develop commercial strategy of the business and build out, you know, the ability for enterprise customers to come and get all the value from these models that we think they can get. So we're really excited about the future. We got hugely exciting pipeline for this year with new modalities and video models and wonderful things and trying to solve images for once and for all and get the kind of general value and value proposition correct for customers. So it's a really exciting time and very honored to be part of it. >> It's great to see some of your customers doing so well out there. Congratulations to your team. Appreciate that. Aidan, let's get into what you guys do. What does Cohere do? What are you excited about right now? >> Yeah, so Cohere builds large language models, which are the backbone of applications like ChatGPT and GPT-3. We're extremely focused on solving the issues with adoption for enterprise. So it's great that you can make a super flashy demo for consumers, but it takes a lot to actually get it into billion user products and large global enterprises. So about six months ago, we released our command models, which are some of the best that exist for large language models. And in December, we released our multilingual text understanding models and that's on over a hundred different languages and it's trained on, you know, authentic data directly from native speakers. And so we're super excited to continue pushing this into enterprise and solving those barriers for adoption, making this transformation a reality. >> Just real quick, while I got you there on the new products coming out. Where are we in the progress? People see some of the new stuff out there right now. There's so much more headroom. Can you just scope out in your mind what that looks like? Like from a headroom standpoint? Okay, we see ChatGPT. "Oh yeah, it writes my papers for me, does some homework for me." I mean okay, yawn, maybe people say that, (Aidan chuckles) people excited or people are blown away. I mean, it's helped theCUBE out, it helps me, you know, feed up a little bit from my write-ups but it's not always perfect. >> Yeah, at the moment it's like a writing assistant, right? And it's still super early in the technologies trajectory. I think it's fascinating and it's interesting but its impact is still really limited. I think in the next year, like within the next eight months, we're going to see some major changes. You've already seen the very first hints of that with stuff like Bing Chat, where you augment these dialogue models with an external knowledge base. So now the models can be kept up to date to the millisecond, right? Because they can search the web and they can see events that happened a millisecond ago. But that's still limited in the sense that when you ask the question, what can these models actually do? Well they can just write text back at you. That's the extent of what they can do. And so the real project, the real effort, that I think we're all working towards is actually taking action. So what happens when you give these models the ability to use tools, to use APIs? What can they do when they can actually affect change out in the real world, beyond just streaming text back at the user? I think that's the really exciting piece. >> Okay, so I wanted to tee that up early in the segment 'cause I want to get into the customer applications. We're seeing early adopters come in, using the technology because they have a lot of data, they have a lot of large language model opportunities and then there's a big fast follower wave coming behind it. I call that the people who are going to jump in the pool early and get into it. They might not be advanced. Can you guys share what customer applications are being used with large language and vision models today and how they're using it to transform on the early adopter side, and how is that a tell sign of what's to come? >> You know, one of the things we have been seeing both with the text models that Aidan talked about as well as the vision models that stability.ai does, Tom, is customers are really using it to change the way you interact with information. You know, one example of a customer that we have, is someone who's kind of using that to query customer conversations and ask questions like, you know, "What was the customer issue? How did we solve it?" And trying to get those kinds of insights that was previously much harder to do. And then of course software is a big area. You know, generating software, making that, you know, just deploying it in production. Those have been really big areas that we have seen customers start to do. You know, looking at documentation, like instead of you know, searching for stuff and so on, you know, you just have an interactive way, in which you can just look at the documentation for a product. You know, all of this goes to where we need to take the technology. One of which is, you know, the models have to be there but they have to work reliably in a production setting at scale, with privacy, with security, and you know, making sure all of this is happening, is going to be really key. That is what, you know, we at AWS are looking to do, which is work with partners like stability and others and in the open source and really take all of these and make them available at scale to customers, where they work reliably. >> Tom, Aidan, what's your thoughts on this? Where are customers landing on this first use cases or set of low-hanging fruit use cases or applications? >> Yeah, so I think like the first group of adopters that really found product market fit were the copywriting companies. So one great example of that is HyperWrite. Another one is Jasper. And so for Cohere, that's the tip of the iceberg, like there's a very long tail of usage from a bunch of different applications. HyperWrite is one of our customers, they help beat writer's block by drafting blog posts, emails, and marketing copy. We also have a global audio streaming platform, which is using us the power of search engine that can comb through podcast transcripts, in a bunch of different languages. Then a global apparel brand, which is using us to transform how they interact with their customers through a virtual assistant, two dozen global news outlets who are using us for news summarization. So really like, these large language models, they can be deployed all over the place into every single industry sector, language is everywhere. It's hard to think of any company on Earth that doesn't use language. So it's, very, very- >> We're doing it right now. We got the language coming in. >> Exactly. >> We'll transcribe this puppy. All right. Tom, on your side, what do you see the- >> Yeah, we're seeing some amazing applications of it and you know, I guess that's partly been, because of the growth in the open source community and some of these applications have come from there that are then triggering this secondary wave of innovation, which is coming a lot from, you know, controllability and explainability of the model. But we've got companies like, you know, Jasper, which Aidan mentioned, who are using stable diffusion for image generation in block creation, content creation. We've got Lensa, you know, which exploded, and is built on top of stable diffusion for fine tuning so people can bring themselves and their pets and you know, everything into the models. So we've now got fine tuned stable diffusion at scale, which is democratized, you know, that process, which is really fun to see your Lensa, you know, exploded. You know, I think it was the largest growing app in the App Store at one point. And lots of other examples like NightCafe and Lexica and Playground. So seeing lots of cool applications. >> So much applications, we'll probably be a customer for all you guys. We'll definitely talk after. But the challenges are there for people adopting, they want to get into what you guys see as the challenges that turn into opportunities. How do you see the customers adopting generative AI applications? For example, we have massive amounts of transcripts, timed up to all the videos. I don't even know what to do. Do I just, do I code my API there. So, everyone has this problem, every vertical has these use cases. What are the challenges for people getting into this and adopting these applications? Is it figuring out what to do first? Or is it a technical setup? Do they stand up stuff, they just go to Amazon? What do you guys see as the challenges? >> I think, you know, the first thing is coming up with where you think you're going to reimagine your customer experience by using generative AI. You know, we talked about Ada, and Tom talked about a number of these ones and you know, you pick up one or two of these, to get that robust. And then once you have them, you know, we have models and we'll have more models on AWS, these large language models that Aidan was talking about. Then you go in and start using these models and testing them out and seeing whether they fit in use case or not. In many situations, like you said, John, our customers want to say, "You know, I know you've trained these models on a lot of publicly available data, but I want to be able to customize it for my use cases. Because, you know, there's some knowledge that I have created and I want to be able to use that." And then in many cases, and I think Aidan mentioned this. You know, you need these models to be up to date. Like you can't have it staying. And in those cases, you augmented with a knowledge base, you know you have to make sure that these models are not hallucinating. And so you need to be able to do the right kind of responsible AI checks. So, you know, you start with a particular use case, and there are a lot of them. Then, you know, you can come to AWS, and then look at one of the many models we have and you know, we are going to have more models for other modalities as well. And then, you know, play around with the models. We have a playground kind of thing where you can test these models on some data and then you can probably, you will probably want to bring your own data, customize it to your own needs, do some of the testing to make sure that the model is giving the right output and then just deploy it. And you know, we have a lot of tools. >> Yeah. >> To make this easy for our customers. >> How should people think about large language models? Because do they think about it as something that they tap into with their IP or their data? Or is it a large language model that they apply into their system? Is the interface that way? What's the interaction look like? >> In many situations, you can use these models out of the box. But in typical, in most of the other situations, you will want to customize it with your own data or with your own expectations. So the typical use case would be, you know, these are models are exposed through APIs. So the typical use case would be, you know you're using these APIs a little bit for testing and getting familiar and then there will be an API that will allow you to train this model further on your data. So you use that AI, you know, make sure you augmented the knowledge base. So then you use those APIs to customize the model and then just deploy it in an application. You know, like Tom was mentioning, a number of companies that are using these models. So once you have it, then you know, you again, use an endpoint API and use it in an application. >> All right, I love the example. I want to ask Tom and Aidan, because like most my experience with Amazon Web Service in 2007, I would stand up in EC2, put my code on there, play around, if it didn't work out, I'd shut it down. Is that a similar dynamic we're going to see with the machine learning where developers just kind of log in and stand up infrastructure and play around and then have a cloud-like experience? >> So I can go first. So I mean, we obviously, with AWS working really closely with the SageMaker team, do fantastic platform there for ML training and inference. And you know, going back to your point earlier, you know, where the data is, is hugely important for companies. Many companies bringing their models to their data in AWS on-premise for them is hugely important. Having the models to be, you know, open sources, makes them explainable and transparent to the adopters of those models. So, you know, we are really excited to work with the SageMaker team over the coming year to bring companies to that platform and make the most of our models. >> Aidan, what's your take on developers? Do they just need to have a team in place, if we want to interface with you guys? Let's say, can they start learning? What do they got to do to set up? >> Yeah, so I think for Cohere, our product makes it much, much easier to people, for people to get started and start building, it solves a lot of the productionization problems. But of course with SageMaker, like Tom was saying, I think that lowers a barrier even further because it solves problems like data privacy. So I want to underline what Bratin was saying earlier around when you're fine tuning or when you're using these models, you don't want your data being incorporated into someone else's model. You don't want it being used for training elsewhere. And so the ability to solve for enterprises, that data privacy and that security guarantee has been hugely important for Cohere, and that's very easy to do through SageMaker. >> Yeah. >> But the barriers for using this technology are coming down super quickly. And so for developers, it's just becoming completely intuitive. I love this, there's this quote from Andrej Karpathy. He was saying like, "It really wasn't on my 2022 list of things to happen that English would become, you know, the most popular programming language." And so the barrier is coming down- >> Yeah. >> Super quickly and it's exciting to see. >> It's going to be awesome for all the companies here, and then we'll do more, we're probably going to see explosion of startups, already seeing that, the maps, ecosystem maps, the landscape maps are happening. So this is happening and I'm convinced it's not yesterday's chat bot, it's not yesterday's AI Ops. It's a whole another ballgame. So I have to ask you guys for the final question before we kick off the company's showcasing here. How do you guys gauge success of generative AI applications? Is there a lens to look through and say, okay, how do I see success? It could be just getting a win or is it a bigger picture? Bratin we'll start with you. How do you gauge success for generative AI? >> You know, ultimately it's about bringing business value to our customers. And making sure that those customers are able to reimagine their experiences by using generative AI. Now the way to get their ease, of course to deploy those models in a safe, effective manner, and ensuring that all of the robustness and the security guarantees and the privacy guarantees are all there. And we want to make sure that this transitions from something that's great demos to actual at scale products, which means making them work reliably all of the time not just some of the time. >> Tom, what's your gauge for success? >> Look, I think this, we're seeing a completely new form of ways to interact with data, to make data intelligent, and directly to bring in new revenue streams into business. So if businesses can use our models to leverage that and generate completely new revenue streams and ultimately bring incredible new value to their customers, then that's fantastic. And we hope we can power that revolution. >> Aidan, what's your take? >> Yeah, reiterating Bratin and Tom's point, I think that value in the enterprise and value in market is like a huge, you know, it's the goal that we're striving towards. I also think that, you know, the value to consumers and actual users and the transformation of the surface area of technology to create experiences like ChatGPT that are magical and it's the first time in human history we've been able to talk to something compelling that's not a human. I think that in itself is just extraordinary and so exciting to see. >> It really brings up a whole another category of markets. B2B, B2C, it's B2D, business to developer. Because I think this is kind of the big trend the consumers have to win. The developers coding the apps, it's a whole another sea change. Reminds me everyone use the "Moneyball" movie as example during the big data wave. Then you know, the value of data. There's a scene in "Moneyball" at the end, where Billy Beane's getting the offer from the Red Sox, then the owner says to the Red Sox, "If every team's not rebuilding their teams based upon your model, there'll be dinosaurs." I think that's the same with AI here. Every company will have to need to think about their business model and how they operate with AI. So it'll be a great run. >> Completely Agree >> It'll be a great run. >> Yeah. >> Aidan, Tom, thank you so much for sharing about your experiences at your companies and congratulations on your success and it's just the beginning. And Bratin, thanks for coming on representing AWS. And thank you, appreciate for what you do. Thank you. >> Thank you, John. Thank you, Aidan. >> Thank you John. >> Thanks so much. >> Okay, let's kick off season three, episode one. I'm John Furrier, your host. Thanks for watching. (light airy music)

(soft electronic music) >> Hey everyone, welcome to this CUBE conversation as part of the AWS Startup Showcase, season three, episode one, featuring Astronomer. I'm your host, Lisa Martin. I'm in the CUBE's Palo Alto Studios, and today excited to be joined by a couple of guests, a couple of co-founders from Astronomer. Viraj Parekh is with us, as is Paola Peraza-Calderon. Thanks guys so much for joining us. Excited to dig into Astronomer. >> Thank you so much for having us. >> Yeah, thanks for having us. >> Yeah, and we're going to be talking about the role of data orchestration. Paola, let's go ahead and start with you. Give the audience that understanding, that context about Astronomer and what it is that you guys do. >> Mm-hmm. Yeah, absolutely. So, Astronomer is a, you know, we're a technology and software company for modern data orchestration, as you said, and we're the driving force behind Apache Airflow. The Open Source Workflow Management tool that's since been adopted by thousands and thousands of users, and we'll dig into this a little bit more. But, by data orchestration, we mean data pipeline, so generally speaking, getting data from one place to another, transforming it, running it on a schedule, and overall just building a central system that tangibly connects your entire ecosystem of data services, right. So what, that's Redshift, Snowflake, DVT, et cetera. And so tangibly, we build, we at Astronomer here build products powered by Apache Airflow for data teams and for data practitioners, so that they don't have to. So, we sell to data engineers, data scientists, data admins, and we really spend our time doing three things. So, the first is that we build Astro, our flagship cloud service that we'll talk more on. But here, we're really building experiences that make it easier for data practitioners to author, run, and scale their data pipeline footprint on the cloud. And then, we also contribute to Apache Airflow as an open source project and community. So, we cultivate the community of humans, and we also put out open source developer tools that actually make it easier for individual data practitioners to be productive in their day-to-day jobs, whether or not they actually use our product and and pay us money or not. And then of course, we also have professional services and education and all of these things around our commercial products that enable folks to use our products and use Airflow as effectively as possible. So yeah, super, super happy with everything we've done and hopefully that gives you an idea of where we're starting. >> Awesome, so when you're talking with those, Paola, those data engineers, those data scientists, how do you define data orchestration and what does it mean to them? >> Yeah, yeah, it's a good question. So, you know, if you Google data orchestration you're going to get something about an automated process for organizing silo data and making it accessible for processing and analysis. But, to your question, what does that actually mean, you know? So, if you look at it from a customer's perspective, we can share a little bit about how we at Astronomer actually do data orchestration ourselves and the problems that it solves for us. So, as many other companies out in the world do, we at Astronomer need to monitor how our own customers use our products, right? And so, we have a weekly meeting, for example, that goes through a dashboard and a dashboarding tool called Sigma where we see the number of monthly customers and how they're engaging with our product. But, to actually do that, you know, we have to use data from our application database, for example, that has behavioral data on what they're actually doing in our product. We also have data from third party API tools, like Salesforce and HubSpot, and other ways in which our customer, we actually engage with our customers and their behavior. And so, our data team internally at Astronomer uses a bunch of tools to transform and use that data, right? So, we use FiveTran, for example, to ingest. We use Snowflake as our data warehouse. We use other tools for data transformations. And even, if we at Astronomer don't do this, you can imagine a data team also using tools like, Monte Carlo for data quality, or Hightouch for Reverse ETL, or things like that. And, I think the point here is that data teams, you know, that are building data-driven organizations have a plethora of tooling to both ingest the right data and come up with the right interfaces to transform and actually, interact with that data. And so, that movement and sort of synchronization of data across your ecosystem is exactly what data orchestration is responsible for. Historically, I think, and Raj will talk more about this, historically, schedulers like KRON and Oozie or Control-M have taken a role here, but we think that Apache Airflow has sort of risen over the past few years as the defacto industry standard for writing data pipelines that do tasks, that do data jobs that interact with that ecosystem of tools in your organization. And so, beyond that sort of data pipeline unit, I think where we see it is that data acquisition is not only writing those data pipelines that move your data, but it's also all the things around it, right, so, CI/CD tool and Secrets Management, et cetera. So, a long-winded answer here, but I think that's how we talk about it here at Astronomer and how we're building our products. >> Excellent. Great context, Paola. Thank you. Viraj, let's bring you into the conversation. Every company these days has to be a data company, right? They've got to be a software company- >> Mm-hmm. >> whether it's my bank or my grocery store. So, how are companies actually doing data orchestration today, Viraj? >> Yeah, it's a great question. So, I think one thing to think about is like, on one hand, you know, data orchestration is kind of a new category that we're helping define, but on the other hand, it's something that companies have been doing forever, right? You need to get data moving to use it, you know. You've got it all in place, aggregate it, cleaning it, et cetera. So, when you look at what companies out there are doing, right. Sometimes, if you're a more kind of born in the cloud company, as we say, you'll adopt all these cloud native tooling things your cloud provider gives you. If you're a bank or another sort of institution like that, you know, you're probably juggling an even wider variety of tools. You're thinking about a cloud migration. You might have things like Kron running in one place, Uzi running somewhere else, Informatics running somewhere else, while you're also trying to move all your workloads to the cloud. So, there's quite a large spectrum of what the current state is for companies. And then, kind of like Paola was saying, Apache Airflow started in 2014, and it was actually started by Airbnb, and they put out this blog post that was like, "Hey here's how we use Apache Airflow to orchestrate our data across all their sources." And really since then, right, it's almost been a decade since then, Airflow emerged as the open source standard, and there's companies of all sorts using it. And, it's really used to tie all these tools together, especially as that number of tools increases, companies move to hybrid cloud, hybrid multi-cloud strategies, and so on and so forth. But you know, what we found is that if you go to any company, especially a larger one and you say like, "Hey, how are you doing data orchestration?" They'll probably say something like, "Well, I have five data teams, so I have eight different ways I do data orchestration." Right. This idea of data orchestration's been there but the right way to do it, kind of all the abstractions you need, the way your teams need to work together, and so on and so forth, hasn't really emerged just yet, right? It's such a quick moving space that companies have to combine what they were doing before with what their new business initiatives are today. So, you know, what we really believe here at Astronomer is Airflow is the core of how you solve data orchestration for any sort of use case, but it's not everything. You know, it needs a little more. And, that's really where our commercial product, Astro comes in, where we've built, not only the most tried and tested airflow experience out there. We do employ a majority of the Airflow Core Committers, right? So, we're kind of really deep in the project. We've also built the right things around developer tooling, observability, and reliability for customers to really rely on Astro as the heart of the way they do data orchestration, and kind of think of it as the foundational layer that helps tie together all the different tools, practices and teams large companies have to do today. >> That foundational layer is absolutely critical. You've both mentioned open source software. Paola, I want to go back to you, and just give the audience an understanding of how open source really plays into Astronomer's mission as a company, and into the technologies like Astro. >> Mm-hmm. Yeah, absolutely. I mean, we, so we at Astronomers started using Airflow and actually building our products because Airflow is open source and we were our own customers at the beginning of our company journey. And, I think the open source community is at the core of everything we do. You know, without that open source community and culture, I think, you know, we have less of a business, and so, we're super invested in continuing to cultivate and grow that. And, I think there's a couple sort of concrete ways in which we do this that personally make me really excited to do my own job. You know, for one, we do things like we organize meetups and we sponsor the Airflow Summit and there's these sort of baseline community efforts that I think are really important and that reminds you, hey, there just humans trying to do their jobs and learn and use both our technology and things that are out there and contribute to it. So, making it easier to contribute to Airflow, for example, is another one of our efforts. As Viraj mentioned, we also employ, you know, engineers internally who are on our team whose full-time job is to make the open source project better. Again, regardless of whether or not you're a customer of ours or not, we want to make sure that we continue to cultivate the Airflow project in and of itself. And, we're also building developer tooling that might not be a part of the Apache Open Source project, but is still open source. So, we have repositories in our own sort of GitHub organization, for example, with tools that individual data practitioners, again customers are not, can use to make them be more productive in their day-to-day jobs with Airflow writing Dags for the most common use cases out there. The last thing I'll say is how important I think we've found it to build sort of educational resources and documentation and best practices. Airflow can be complex. It's been around for a long time. There's a lot of really, really rich feature sets. And so, how do we enable folks to actually use those? And that comes in, you know, things like webinars, and best practices, and courses and curriculum that are free and accessible and open to the community are just some of the ways in which I think we're continuing to invest in that open source community over the next year and beyond. >> That's awesome. It sounds like open source is really core, not only to the mission, but really to the heart of the organization. Viraj, I want to go back to you and really try to understand how does Astronomer fit into the wider modern data stack and ecosystem? Like what does that look like for customers? >> Yeah, yeah. So, both in the open source and with our commercial customers, right? Folks everywhere are trying to tie together a huge variety of tools in order to start making sense of their data. And you know, I kind of think of it almost like as like a pyramid, right? At the base level, you need things like data reliability, data, sorry, data freshness, data availability, and so on and so forth, right? You just need your data to be there. (coughs) I'm sorry. You just need your data to be there, and you need to make it predictable when it's going to be there. You need to make sure it's kind of correct at the highest level, some quality checks, and so on and so forth. And oftentimes, that kind of takes the case of ELT or ETL use cases, right? Taking data from somewhere and moving it somewhere else, usually into some sort of analytics destination. And, that's really what businesses can do to just power the core parts of getting insights into how their business is going, right? How much revenue did I had? What's in my pipeline, salesforce, and so on and so forth. Once that kind of base foundation is there and people can get the data they need, how they need it, it really opens up a lot for what customers can do. You know, I think one of the trendier things out there right now is MLOps, and how do companies actually put machine learning into production? Well, when you think about it you kind of have to squint at it, right? Like, machine learning pipelines are really just any other data pipeline. They just have a certain set of needs that might not not be applicable to ELT pipelines. And, when you kind of have a common layer to tie together all the ways data can move through your organization, that's really what we're trying to make it so companies can do. And, that happens in financial services where, you know, we have some customers who take app data coming from their mobile apps, and actually run it through their fraud detection services to make sure that all the activity is not fraudulent. We have customers that will run sports betting models on our platform where they'll take data from a bunch of public APIs around different sporting events that are happening, transform all of that in a way their data scientist can build models with it, and then actually bet on sports based on that output. You know, one of my favorite use cases I like to talk about that we saw in the open source is we had there was one company whose their business was to deliver blood transfusions via drone into remote parts of the world. And, it was really cool because they took all this data from all sorts of places, right? Kind of orchestrated all the aggregation and cleaning and analysis that happened had to happen via airflow and the end product would be a drone being shot out into a real remote part of the world to actually give somebody blood who needed it there. Because it turns out for certain parts of the world, the easiest way to deliver blood to them is via drone and not via some other, some other thing. So, these kind of, all the things people do with the modern data stack is absolutely incredible, right? Like you were saying, every company's trying to be a data-driven company. What really energizes me is knowing that like, for all those best, super great tools out there that power a business, we get to be the connective tissue, or the, almost like the electricity that kind of ropes them all together and makes so people can actually do what they need to do. >> Right. Phenomenal use cases that you just described, Raj. I mean, just the variety alone of what you guys are able to do and impact is so cool. So Paola, when you're with those data engineers, those data scientists, and customer conversations, what's your pitch? Why use Astro? >> Mm-hmm. Yeah, yeah, it's a good question. And honestly, to piggyback off of Viraj, there's so many. I think what keeps me so energized is how mission critical both our product and data orchestration is, and those use cases really are incredible and we work with customers of all shapes and sizes. But, to answer your question, right, so why use Astra? Why use our commercial products? There's so many people using open source, why pay for something more than that? So, you know, the baseline for our business really is that Airflow has grown exponentially over the last five years, and like we said has become an industry standard that we're confident there's a huge opportunity for us as a company and as a team. But, we also strongly believe that being great at running Airflow, you know, doesn't make you a successful company at what you do. What makes you a successful company at what you do is building great products and solving problems and solving pin points of your own customers, right? And, that differentiating value isn't being amazing at running Airflow. That should be our job. And so, we want to abstract those customers from meaning to do things like manage Kubernetes infrastructure that you need to run Airflow, and then hiring someone full-time to go do that. Which can be hard, but again doesn't add differentiating value to your team, or to your product, or to your customers. So, folks to get away from managing that infrastructure sort of a base, a base layer. Folks who are looking for differentiating features that make their team more productive and allows them to spend less time tweaking Airflow configurations and more time working with the data that they're getting from their business. For help, getting, staying up with Airflow releases. There's a ton of, we've actually been pretty quick to come out with new Airflow features and releases, and actually just keeping up with that feature set and working strategically with a partner to help you make the most out of those feature sets is a key part of it. And, really it's, especially if you're an organization who currently is committed to using Airflow, you likely have a lot of Airflow environments across your organization. And, being able to see those Airflow environments in a single place and being able to enable your data practitioners to create Airflow environments with a click of a button, and then use, for example, our command line to develop your Airflow Dags locally and push them up to our product, and use all of the sort of testing and monitoring and observability that we have on top of our product is such a key. It sounds so simple, especially if you use Airflow, but really those things are, you know, baseline value props that we have for the customers that continue to be excited to work with us. And of course, I think we can go beyond that and there's, we have ambitions to add whole, a whole bunch of features and expand into different types of personas. >> Right? >> But really our main value prop is for companies who are committed to Airflow and want to abstract themselves and make use of some of the differentiating features that we now have at Astronomer. >> Got it. Awesome. >> Thank you. One thing, one thing I'll add to that, Paola, and I think you did a good job of saying is because every company's trying to be a data company, companies are at different parts of their journey along that, right? And we want to meet customers where they are, and take them through it to where they want to go. So, on one end you have folks who are like, "Hey, we're just building a data team here. We have a new initiative. We heard about Airflow. How do you help us out?" On the farther end, you know, we have some customers that have been using Airflow for five plus years and they're like, "Hey, this is awesome. We have 10 more teams we want to bring on. How can you help with this? How can we do more stuff in the open source with you? How can we tell our story together?" And, it's all about kind of taking this vast community of data users everywhere, seeing where they're at, and saying like, "Hey, Astro and Airflow can take you to the next place that you want to go." >> Which is incredibly- >> Mm-hmm. >> and you bring up a great point, Viraj, that every company is somewhere in a different place on that journey. And it's, and it's complex. But it sounds to me like a lot of what you're doing is really stripping away a lot of the complexity, really enabling folks to use their data as quickly as possible, so that it's relevant and they can serve up, you know, the right products and services to whoever wants what. Really incredibly important. We're almost out of time, but I'd love to get both of your perspectives on what's next for Astronomer. You give us a a great overview of what the company's doing, the value in it for customers. Paola, from your lens as one of the co-founders, what's next? >> Yeah, I mean, I think we'll continue to, I think cultivate in that open source community. I think we'll continue to build products that are open sourced as part of our ecosystem. I also think that we'll continue to build products that actually make Airflow, and getting started with Airflow, more accessible. So, sort of lowering that barrier to entry to our products, whether that's price wise or infrastructure requirement wise. I think making it easier for folks to get started and get their hands on our product is super important for us this year. And really it's about, I think, you know, for us, it's really about focused execution this year and all of the sort of core principles that we've been talking about. And continuing to invest in all of the things around our product that again, enable teams to use Airflow more effectively and efficiently. >> And that efficiency piece is, everybody needs that. Last question, Viraj, for you. What do you see in terms of the next year for Astronomer and for your role? >> Yeah, you know, I think Paola did a really good job of laying it out. So it's, it's really hard to disagree with her on anything, right? I think executing is definitely the most important thing. My own personal bias on that is I think more than ever it's important to really galvanize the community around airflow. So, we're going to be focusing on that a lot. We want to make it easier for our users to get get our product into their hands, be that open source users or commercial users. And last, but certainly not least, is we're also really excited about Data Lineage and this other open source project in our umbrella called Open Lineage to make it so that there's a standard way for users to get lineage out of different systems that they use. When we think about what's in store for data lineage and needing to audit the way automated decisions are being made. You know, I think that's just such an important thing that companies are really just starting with, and I don't think there's a solution that's emerged that kind of ties it all together. So, we think that as we kind of grow the role of Airflow, right, we can also make it so that we're helping solve, we're helping customers solve their lineage problems all in Astro, which is our kind of the best of both worlds for us. >> Awesome. I can definitely feel and hear the enthusiasm and the passion that you both bring to Astronomer, to your customers, to your team. I love it. We could keep talking more and more, so you're going to have to come back. (laughing) Viraj, Paola, thank you so much for joining me today on this showcase conversation. We really appreciate your insights and all the context that you provided about Astronomer. >> Thank you so much for having us. >> My pleasure. For my guests, I'm Lisa Martin. You're watching this Cube conversation. (soft electronic music)

(upbeat electronic music) >> Hello everyone, welcome to this CUBE conversation here from the studios in the CUBE in Palo Alto, California. I'm John Furrier, your host. We're featuring a startup, Astronomer. Astronomer.io is the URL, check it out. And we're going to have a great conversation around one of the most important topics hitting the industry, and that is the future of machine learning and AI, and the data that powers it underneath it. There's a lot of things that need to get done, and we're excited to have some of the co-founders of Astronomer here. Viraj Parekh, who is co-founder of Astronomer, and Paola Peraza Calderon, another co-founder, both with Astronomer. Thanks for coming on. First of all, how many co-founders do you guys have? >> You know, I think the answer's around six or seven. I forget the exact, but there's really been a lot of people around the table who've worked very hard to get this company to the point that it's at. We have long ways to go, right? But there's been a lot of people involved that have been absolutely necessary for the path we've been on so far. >> Thanks for that, Viraj, appreciate that. The first question I want to get out on the table, and then we'll get into some of the details, is take a minute to explain what you guys are doing. How did you guys get here? Obviously, multiple co-founders, sounds like a great project. The timing couldn't have been better. ChatGPT has essentially done so much public relations for the AI industry to kind of highlight this shift that's happening. It's real, we've been chronicalizing, take a minute to explain what you guys do. >> Yeah, sure, we can get started. So, yeah, when Viraj and I joined Astronomer in 2017, we really wanted to build a business around data, and we were using an open source project called Apache Airflow that we were just using sort of as customers ourselves. And over time, we realized that there was actually a market for companies who use Apache Airflow, which is a data pipeline management tool, which we'll get into, and that running Airflow is actually quite challenging, and that there's a big opportunity for us to create a set of commercial products and an opportunity to grow that open source community and actually build a company around that. So the crux of what we do is help companies run data pipelines with Apache Airflow. And certainly we've grown in our ambitions beyond that, but that's sort of the crux of what we do for folks. >> You know, data orchestration, data management has always been a big item in the old classic data infrastructure. But with AI, you're seeing a lot more emphasis on scale, tuning, training. Data orchestration is the center of the value proposition, when you're looking at coordinating resources, it's one of the most important things. Can you guys explain what data orchestration entails? What does it mean? Take us through the definition of what data orchestration entails. >> Yeah, for sure. I can take this one, and Viraj, feel free to jump in. So if you google data orchestration, here's what you're going to get. You're going to get something that says, "Data orchestration is the automated process" "for organizing silo data from numerous" "data storage points, standardizing it," "and making it accessible and prepared for data analysis." And you say, "Okay, but what does that actually mean," right, and so let's give sort of an an example. So let's say you're a business and you have sort of the following basic asks of your data team, right? Okay, give me a dashboard in Sigma, for example, for the number of customers or monthly active users, and then make sure that that gets updated on an hourly basis. And then number two, a consistent list of active customers that I have in HubSpot so that I can send them a monthly product newsletter, right? Two very basic asks for all sorts of companies and organizations. And when that data team, which has data engineers, data scientists, ML engineers, data analysts get that request, they're looking at an ecosystem of data sources that can help them get there, right? And that includes application databases, for example, that actually have in product user behavior and third party APIs from tools that the company uses that also has different attributes and qualities of those customers or users. And that data team needs to use tools like Fivetran to ingest data, a data warehouse, like Snowflake or Databricks to actually store that data and do analysis on top of it, a tool like DBT to do transformations and make sure that data is standardized in the way that it needs to be, a tool like Hightouch for reverse ETL. I mean, we could go on and on. There's so many partners of ours in this industry that are doing really, really exciting and critical things for those data movements. And the whole point here is that data teams have this plethora of tooling that they use to both ingest the right data and come up with the right interfaces to transform and interact with that data. And data orchestration, in our view, is really the heartbeat of all of those processes, right? And tangibly the unit of data orchestration is a data pipeline, a set of tasks or jobs that each do something with data over time and eventually run that on a schedule to make sure that those things are happening continuously as time moves on and the company advances. And so, for us, we're building a business around Apache Airflow, which is a workflow management tool that allows you to author, run, and monitor data pipelines. And so when we talk about data orchestration, we talk about sort of two things. One is that crux of data pipelines that, like I said, connect that large ecosystem of data tooling in your company. But number two, it's not just that data pipeline that needs to run every day, right? And Viraj will probably touch on this as we talk more about Astronomer and our value prop on top of Airflow. But then it's all the things that you need to actually run data and production and make sure that it's trustworthy, right? So it's actually not just that you're running things on a schedule, but it's also things like CICD tooling, secure secrets management, user permissions, monitoring, data lineage, documentation, things that enable other personas in your data team to actually use those tools. So long-winded way of saying that it's the heartbeat, we think, of of the data ecosystem, and certainly goes beyond scheduling, but again, data pipelines are really at the center of it. >> One of the things that jumped out, Viraj, if you can get into this, I'd like to hear more about how you guys look at all those little tools that are out. You mentioned a variety of things. You look at the data infrastructure, it's not just one stack. You've got an analytic stack, you've got a realtime stack, you've got a data lake stack, you got an AI stack potentially. I mean you have these stacks now emerging in the data world that are fundamental, that were once served by either a full package, old school software, and then a bunch of point solution. You mentioned Fivetran there, I would say in the analytics stack. Then you got S3, they're on the data lake stack. So all these things are kind of munged together. >> Yeah. >> How do you guys fit into that world? You make it easier, or like, what's the deal? >> Great question, right? And you know, I think that one of the biggest things we've found in working with customers over the last however many years is that if a data team is using a bunch of tools to get what they need done, and the number of tools they're using is growing exponentially and they're kind of roping things together here and there, that's actually a sign of a productive team, not a bad thing, right? It's because that team is moving fast. They have needs that are very specific to them, and they're trying to make something that's exactly tailored to their business. So a lot of times what we find is that customers have some sort of base layer, right? That's kind of like, it might be they're running most of the things in AWS, right? And then on top of that, they'll be using some of the things AWS offers, things like SageMaker, Redshift, whatever, but they also might need things that their cloud can't provide. Something like Fivetran, or Hightouch, those are other tools. And where data orchestration really shines, and something that we've had the pleasure of helping our customers build, is how do you take all those requirements, all those different tools and whip them together into something that fulfills a business need? So that somebody can read a dashboard and trust the number that it says, or somebody can make sure that the right emails go out to their customers. And Airflow serves as this amazing kind of glue between that data stack, right? It's to make it so that for any use case, be it ELT pipelines, or machine learning, or whatever, you need different things to do them, and Airflow helps tie them together in a way that's really specific for a individual business' needs. >> Take a step back and share the journey of what you guys went through as a company startup. So you mentioned Apache, open source. I was just having an interview with a VC, we were talking about foundational models. You got a lot of proprietary and open source development going on. It's almost the iPhone/Android moment in this whole generative space and foundational side. This is kind of important, the open source piece of it. Can you share how you guys started? And I can imagine your customers probably have their hair on fire and are probably building stuff on their own. Are you guys helping them? Take us through, 'cause you guys are on the front end of a big, big wave, and that is to make sense of the chaos, rain it in. Take us through your journey and why this is important. >> Yeah, Paola, I can take a crack at this, then I'll kind of hand it over to you to fill in whatever I miss in details. But you know, like Paola is saying, the heart of our company is open source, because we started using Airflow as an end user and started to say like, "Hey wait a second," "more and more people need this." Airflow, for background, started at Airbnb, and they were actually using that as a foundation for their whole data stack. Kind of how they made it so that they could give you recommendations, and predictions, and all of the processes that needed orchestrated. Airbnb created Airflow, gave it away to the public, and then fast forward a couple years and we're building a company around it, and we're really excited about that. >> That's a beautiful thing. That's exactly why open source is so great. >> Yeah, yeah. And for us, it's really been about watching the community and our customers take these problems, find a solution to those problems, standardize those solutions, and then building on top of that, right? So we're reaching to a point where a lot of our earlier customers who started to just using Airflow to get the base of their BI stack down and their reporting in their ELP infrastructure, they've solved that problem and now they're moving on to things like doing machine learning with their data, because now that they've built that foundation, all the connective tissue for their data arriving on time and being orchestrated correctly is happening, they can build a layer on top of that. And it's just been really, really exciting kind of watching what customers do once they're empowered to pick all the tools that they need, tie them together in the way they need to, and really deliver real value to their business. >> Can you share some of the use cases of these customers? Because I think that's where you're starting to see the innovation. What are some of the companies that you're working with, what are they doing? >> Viraj, I'll let you take that one too. (group laughs) >> So you know, a lot of it is... It goes across the gamut, right? Because it doesn't matter what you are, what you're doing with data, it needs to be orchestrated. So there's a lot of customers using us for their ETL and ELT reporting, right? Just getting data from other disparate sources into one place and then building on top of that. Be it building dashboards, answering questions for the business, building other data products and so on and so forth. From there, these use cases evolve a lot. You do see folks doing things like fraud detection, because Airflow's orchestrating how transactions go, transactions get analyzed. They do things like analyzing marketing spend to see where your highest ROI is. And then you kind of can't not talk about all of the machine learning that goes on, right? Where customers are taking data about their own customers, kind of analyze and aggregating that at scale, and trying to automate decision making processes. So it goes from your most basic, what we call data plumbing, right? Just to make sure data's moving as needed, all the ways to your more exciting expansive use cases around automated decision making and machine learning. >> And I'd say, I mean, I'd say that's one of the things that I think gets me most excited about our future, is how critical Airflow is to all of those processes, and I think when you know a tool is valuable is when something goes wrong and one of those critical processes doesn't work. And we know that our system is so mission critical to answering basic questions about your business and the growth of your company for so many organizations that we work with. So it's, I think, one of the things that gets Viraj and I and the rest of our company up every single morning is knowing how important the work that we do for all of those use cases across industries, across company sizes, and it's really quite energizing. >> It was such a big focus this year at AWS re:Invent, the role of data. And I think one of the things that's exciting about the open AI and all the movement towards large language models is that you can integrate data into these models from outside. So you're starting to see the integration easier to deal with. Still a lot of plumbing issues. So a lot of things happening. So I have to ask you guys, what is the state of the data orchestration area? Is it ready for disruption? Has it already been disrupted? Would you categorize it as a new first inning kind of opportunity, or what's the state of the data orchestration area right now? Both technically and from a business model standpoint. How would you guys describe that state of the market? >> Yeah, I mean, I think in a lot of ways, in some ways I think we're category creating. Schedulers have been around for a long time. I released a data presentation sort of on the evolution of going from something like Kron, which I think was built in like the 1970s out of Carnegie Mellon. And that's a long time ago, that's 50 years ago. So sort of like the basic need to schedule and do something with your data on a schedule is not a new concept. But to our point earlier, I think everything that you need around your ecosystem, first of all, the number of data tools and developer tooling that has come out industry has 5X'd over the last 10 years. And so obviously as that ecosystem grows, and grows, and grows, and grows, the need for orchestration only increases. And I think, as Astronomer, I think we... And we work with so many different types of companies, companies that have been around for 50 years, and companies that got started not even 12 months ago. And so I think for us it's trying to, in a ways, category create and adjust sort of what we sell and the value that we can provide for companies all across that journey. There are folks who are just getting started with orchestration, and then there's folks who have such advanced use case, 'cause they're hitting sort of a ceiling and only want to go up from there. And so I think we, as a company, care about both ends of that spectrum, and certainly want to build and continue building products for companies of all sorts, regardless of where they are on the maturity curve of data orchestration. >> That's a really good point, Paola. And I think the other thing to really take into account is it's the companies themselves, but also individuals who have to do their jobs. If you rewind the clock like 5 or 10 years ago, data engineers would be the ones responsible for orchestrating data through their org. But when we look at our customers today, it's not just data engineers anymore. There's data analysts who sit a lot closer to the business, and the data scientists who want to automate things around their models. So this idea that orchestration is this new category is right on the money. And what we're finding is the need for it is spreading to all parts of the data team, naturally where Airflow's emerged as an open source standard and we're hoping to take things to the next level. >> That's awesome. We've been up saying that the data market's kind of like the SRE with servers, right? You're going to need one person to deal with a lot of data, and that's data engineering, and then you're got to have the practitioners, the democratization. Clearly that's coming in what you're seeing. So I have to ask, how do you guys fit in from a value proposition standpoint? What's the pitch that you have to customers, or is it more inbound coming into you guys? Are you guys doing a lot of outreach, customer engagements? I'm sure they're getting a lot of great requirements from customers. What's the current value proposition? How do you guys engage? >> Yeah, I mean, there's so many... Sorry, Viraj, you can jump in. So there's so many companies using Airflow, right? So the baseline is that the open source project that is Airflow that came out of Airbnb, over five years ago at this point, has grown exponentially in users and continues to grow. And so the folks that we sell to primarily are folks who are already committed to using Apache Airflow, need data orchestration in their organization, and just want to do it better, want to do it more efficiently, want to do it without managing that infrastructure. And so our baseline proposition is for those organizations. Now to Viraj's point, obviously I think our ambitions go beyond that, both in terms of the personas that we addressed and going beyond that data engineer, but really it's to start at the baseline, as we continue to grow our our company, it's really making sure that we're adding value to folks using Airflow and help them do so in a better way, in a larger way, in a more efficient way, and that's really the crux of who we sell to. And so to answer your question on, we get a lot of inbound because they're... >> You have a built in audience. (laughs) >> The world that use it. Those are the folks who we talk to and come to our website and chat with us and get value from our content. I mean, the power of the opensource community is really just so, so big, and I think that's also one of the things that makes this job fun. >> And you guys are in a great position. Viraj, you can comment a little, get your reaction. There's been a big successful business model to starting a company around these big projects for a lot of reasons. One is open source is continuing to be great, but there's also supply chain challenges in there. There's also we want to continue more innovation and more code and keeping it free and and flowing. And then there's the commercialization of productizing it, operationalizing it. This is a huge new dynamic, I mean, in the past 5 or so years, 10 years, it's been happening all on CNCF from other areas like Apache, Linux Foundation, they're all implementing this. This is a huge opportunity for entrepreneurs to do this. >> Yeah, yeah. Open source is always going to be core to what we do, because we wouldn't exist without the open source community around us. They are huge in numbers. Oftentimes they're nameless people who are working on making something better in a way that everybody benefits from it. But open source is really hard, especially if you're a company whose core competency is running a business, right? Maybe you're running an e-commerce business, or maybe you're running, I don't know, some sort of like, any sort of business, especially if you're a company running a business, you don't really want to spend your time figuring out how to run open source software. You just want to use it, you want to use the best of it, you want to use the community around it, you want to be able to google something and get answers for it, you want the benefits of open source. You don't have the time or the resources to invest in becoming an expert in open source, right? And I think that dynamic is really what's given companies like us an ability to kind of form businesses around that in the sense that we'll make it so people get the best of both worlds. You'll get this vast open ecosystem that you can build on top of, that you can benefit from, that you can learn from. But you won't have to spend your time doing undifferentiated heavy lifting. You can do things that are just specific to your business. >> It's always been great to see that business model evolve. We used a debate 10 years ago, can there be another Red Hat? And we said, not really the same, but there'll be a lot of little ones that'll grow up to be big soon. Great stuff. Final question, can you guys share the history of the company? The milestones of Astromer's journey in data orchestration? >> Yeah, we could. So yeah, I mean, I think, so Viraj and I have obviously been at Astronomer along with our other founding team and leadership folks for over five years now. And it's been such an incredible journey of learning, of hiring really amazing people, solving, again, mission critical problems for so many types of organizations. We've had some funding that has allowed us to invest in the team that we have and in the software that we have, and that's been really phenomenal. And so that investment, I think, keeps us confident, even despite these sort of macroeconomic conditions that we're finding ourselves in. And so honestly, the milestones for us are focusing on our product, focusing on our customers over the next year, focusing on that market for us that we know can get valuable out of what we do, and making developers' lives better, and growing the open source community and making sure that everything that we're doing makes it easier for folks to get started, to contribute to the project and to feel a part of the community that we're cultivating here. >> You guys raised a little bit of money. How much have you guys raised? >> Don't know what the total is, but it's in the ballpark over $200 million. It feels good to... >> A little bit of capital. Got a little bit of cap to work with there. Great success. I know as a Series C Financing, you guys have been down. So you're up and running, what's next? What are you guys looking to do? What's the big horizon look like for you from a vision standpoint, more hiring, more product, what is some of the key things you're looking at doing? >> Yeah, it's really a little of all of the above, right? Kind of one of the best and worst things about working at earlier stage startups is there's always so much to do and you often have to just kind of figure out a way to get everything done. But really investing our product over the next, at least over the course of our company lifetime. And there's a lot of ways we want to make it more accessible to users, easier to get started with, easier to use, kind of on all areas there. And really, we really want to do more for the community, right, like I was saying, we wouldn't be anything without the large open source community around us. And we want to figure out ways to give back more in more creative ways, in more code driven ways, in more kind of events and everything else that we can keep those folks galvanized and just keep them happy using Airflow. >> Paola, any final words as we close out? >> No, I mean, I'm super excited. I think we'll keep growing the team this year. We've got a couple of offices in the the US, which we're excited about, and a fully global team that will only continue to grow. So Viraj and I are both here in New York, and we're excited to be engaging with our coworkers in person finally, after years of not doing so. We've got a bustling office in San Francisco as well. So growing those teams and continuing to hire all over the world, and really focusing on our product and the open source community is where our heads are at this year. So, excited. >> Congratulations. 200 million in funding, plus. Good runway, put that money in the bank, squirrel it away. It's a good time to kind of get some good interest on it, but still grow. Congratulations on all the work you guys do. We appreciate you and the open source community does, and good luck with the venture, continue to be successful, and we'll see you at the Startup Showcase. >> Thank you. >> Yeah, thanks so much, John. Appreciate it. >> Okay, that's the CUBE Conversation featuring astronomer.io, that's the website. Astronomer is doing well. Multiple rounds of funding, over 200 million in funding. Open source continues to lead the way in innovation. Great business model, good solution for the next gen cloud scale data operations, data stacks that are emerging. I'm John Furrier, your host, thanks for watching. (soft upbeat music)

(soft music) >> Hello everyone, welcome to this Cube conversation here from the studios of theCube in Palo Alto, California. John Furrier, your host. We're featuring a startup, Astronomer, astronomer.io is the url. Check it out. And we're going to have a great conversation around one of the most important topics hitting the industry, and that is the future of machine learning and AI and the data that powers it underneath it. There's a lot of things that need to get done, and we're excited to have some of the co-founders of Astronomer here. Viraj Parekh, who is co-founder and Paola Peraza Calderon, another co-founder, both with Astronomer. Thanks for coming on. First of all, how many co-founders do you guys have? >> You know, I think the answer's around six or seven. I forget the exact, but there's really been a lot of people around the table, who've worked very hard to get this company to the point that it's at. And we have long ways to go, right? But there's been a lot of people involved that are, have been absolutely necessary for the path we've been on so far. >> Thanks for that, Viraj, appreciate that. The first question I want to get out on the table, and then we'll get into some of the details, is take a minute to explain what you guys are doing. How did you guys get here? Obviously, multiple co-founders sounds like a great project. The timing couldn't have been better. ChatGPT has essentially done so much public relations for the AI industry. Kind of highlight this shift that's happening. It's real. We've been chronologicalizing, take a minute to explain what you guys do. >> Yeah, sure. We can get started. So yeah, when Astronomer, when Viraj and I joined Astronomer in 2017, we really wanted to build a business around data and we were using an open source project called Apache Airflow, that we were just using sort of as customers ourselves. And over time, we realized that there was actually a market for companies who use Apache Airflow, which is a data pipeline management tool, which we'll get into. And that running Airflow is actually quite challenging and that there's a lot of, a big opportunity for us to create a set of commercial products and opportunity to grow that open source community and actually build a company around that. So the crux of what we do is help companies run data pipelines with Apache Airflow. And certainly we've grown in our ambitions beyond that, but that's sort of the crux of what we do for folks. >> You know, data orchestration, data management has always been a big item, you know, in the old classic data infrastructure. But with AI you're seeing a lot more emphasis on scale, tuning, training. You know, data orchestration is the center of the value proposition when you're looking at coordinating resources, it's one of the most important things. Could you guys explain what data orchestration entails? What does it mean? Take us through the definition of what data orchestration entails. >> Yeah, for sure. I can take this one and Viraj feel free to jump in. So if you google data orchestration, you know, here's what you're going to get. You're going to get something that says, data orchestration is the automated process for organizing silo data from numerous data storage points to organizing it and making it accessible and prepared for data analysis. And you say, okay, but what does that actually mean, right? And so let's give sort of an example. So let's say you're a business and you have sort of the following basic asks of your data team, right? Hey, give me a dashboard in Sigma, for example, for the number of customers or monthly active users and then make sure that that gets updated on an hourly basis. And then number two, a consistent list of active customers that I have in HubSpot so that I can send them a monthly product newsletter, right? Two very basic asks for all sorts of companies and organizations. And when that data team, which has data engineers, data scientists, ML engineers, data analysts get that request, they're looking at an ecosystem of data sources that can help them get there, right? And that includes application databases, for example, that actually have end product user behavior and third party APIs from tools that the company uses that also has different attributes and qualities of those customers or users. And that data team needs to use tools like Fivetran, to ingest data, a data warehouse like Snowflake or Databricks to actually store that data and do analysis on top of it, a tool like DBT to do transformations and make sure that that data is standardized in the way that it needs to be, a tool like Hightouch for reverse ETL. I mean, we could go on and on. There's so many partners of ours in this industry that are doing really, really exciting and critical things for those data movements. And the whole point here is that, you know, data teams have this plethora of tooling that they use to both ingest the right data and come up with the right interfaces to transform and interact with that data. And data orchestration in our view is really the heartbeat of all of those processes, right? And tangibly the unit of data orchestration, you know, is a data pipeline, a set of tasks or jobs that each do something with data over time and eventually run that on a schedule to make sure that those things are happening continuously as time moves on. And, you know, the company advances. And so, you know, for us, we're building a business around Apache Airflow, which is a workflow management tool that allows you to author, run and monitor data pipelines. And so when we talk about data orchestration, we talk about sort of two things. One is that crux of data pipelines that, like I said, connect that large ecosystem of data tooling in your company. But number two, it's not just that data pipeline that needs to run every day, right? And Viraj will probably touch on this as we talk more about Astronomer and our value prop on top of Airflow. But then it's all the things that you need to actually run data and production and make sure that it's trustworthy, right? So it's actually not just that you're running things on a schedule, but it's also things like CI/CD tooling, right? Secure secrets management, user permissions, monitoring, data lineage, documentation, things that enable other personas in your data team to actually use those tools. So long-winded way of saying that, it's the heartbeat that we think of the data ecosystem and certainly goes beyond scheduling, but again, data pipelines are really at the center of it. >> You know, one of the things that jumped out Viraj, if you can get into this, I'd like to hear more about how you guys look at all those little tools that are out there. You mentioned a variety of things. You know, if you look at the data infrastructure, it's not just one stack. You've got an analytic stack, you've got a realtime stack, you've got a data lake stack, you got an AI stack potentially. I mean you have these stacks now emerging in the data world that are >> Yeah. - >> fundamental, but we're once served by either a full package, old school software, and then a bunch of point solution. You mentioned Fivetran there, I would say in the analytics stack. Then you got, you know, S3, they're on the data lake stack. So all these things are kind of munged together. >> Yeah. >> How do you guys fit into that world? You make it easier or like, what's the deal? >> Great question, right? And you know, I think that one of the biggest things we've found in working with customers over, you know, the last however many years, is that like if a data team is using a bunch of tools to get what they need done and the number of tools they're using is growing exponentially and they're kind of roping things together here and there, that's actually a sign of a productive team, not a bad thing, right? It's because that team is moving fast. They have needs that are very specific to them and they're trying to make something that's exactly tailored to their business. So a lot of times what we find is that customers have like some sort of base layer, right? That's kind of like, you know, it might be they're running most of the things in AWS, right? And then on top of that, they'll be using some of the things AWS offers, you know, things like SageMaker, Redshift, whatever. But they also might need things that their Cloud can't provide, you know, something like Fivetran or Hightouch or anything of those other tools and where data orchestration really shines, right? And something that we've had the pleasure of helping our customers build, is how do you take all those requirements, all those different tools and whip them together into something that fulfills a business need, right? Something that makes it so that somebody can read a dashboard and trust the number that it says or somebody can make sure that the right emails go out to their customers. And Airflow serves as this amazing kind of glue between that data stack, right? It's to make it so that for any use case, be it ELT pipelines or machine learning or whatever, you need different things to do them and Airflow helps tie them together in a way that's really specific for a individual business's needs. >> Take a step back and share the journey of what your guys went through as a company startup. So you mentioned Apache open source, you know, we were just, I was just having an interview with the VC, we were talking about foundational models. You got a lot of proprietary and open source development going on. It's almost the iPhone, Android moment in this whole generative space and foundational side. This is kind of important, the open source piece of it. Can you share how you guys started? And I can imagine your customers probably have their hair on fire and are probably building stuff on their own. How do you guys, are you guys helping them? Take us through, 'cuz you guys are on the front end of a big, big wave and that is to make sense of the chaos, reigning it in. Take us through your journey and why this is important. >> Yeah Paola, I can take a crack at this and then I'll kind of hand it over to you to fill in whatever I miss in details. But you know, like Paola is saying, the heart of our company is open source because we started using Airflow as an end user and started to say like, "Hey wait a second". Like more and more people need this. Airflow, for background, started at Airbnb and they were actually using that as the foundation for their whole data stack. Kind of how they made it so that they could give you recommendations and predictions and all of the processes that need to be or needed to be orchestrated. Airbnb created Airflow, gave it away to the public and then, you know, fast forward a couple years and you know, we're building a company around it and we're really excited about that. >> That's a beautiful thing. That's exactly why open source is so great. >> Yeah, yeah. And for us it's really been about like watching the community and our customers take these problems, find solution to those problems, build standardized solutions, and then building on top of that, right? So we're reaching to a point where a lot of our earlier customers who started to just using Airflow to get the base of their BI stack down and their reporting and their ELP infrastructure, you know, they've solved that problem and now they're moving onto things like doing machine learning with their data, right? Because now that they've built that foundation, all the connective tissue for their data arriving on time and being orchestrated correctly is happening, they can build the layer on top of that. And it's just been really, really exciting kind of watching what customers do once they're empowered to pick all the tools that they need, tie them together in the way they need to, and really deliver real value to their business. >> Can you share some of the use cases of these customers? Because I think that's where you're starting to see the innovation. What are some of the companies that you're working with, what are they doing? >> Raj, I'll let you take that one too. (all laughing) >> Yeah. (all laughing) So you know, a lot of it is, it goes across the gamut, right? Because all doesn't matter what you are, what you're doing with data, it needs to be orchestrated. So there's a lot of customers using us for their ETL and ELT reporting, right? Just getting data from all the disparate sources into one place and then building on top of that, be it building dashboards, answering questions for the business, building other data products and so on and so forth. From there, these use cases evolve a lot. You do see folks doing things like fraud detection because Airflow's orchestrating how transactions go. Transactions get analyzed, they do things like analyzing marketing spend to see where your highest ROI is. And then, you know, you kind of can't not talk about all of the machine learning that goes on, right? Where customers are taking data about their own customers kind of analyze and aggregating that at scale and trying to automate decision making processes. So it goes from your most basic, what we call like data plumbing, right? Just to make sure data's moving as needed. All the ways to your more exciting and sexy use cases around like automated decision making and machine learning. >> And I'd say, I mean, I'd say that's one of the things that I think gets me most excited about our future is how critical Airflow is to all of those processes, you know? And I think when, you know, you know a tool is valuable is when something goes wrong and one of those critical processes doesn't work. And we know that our system is so mission critical to answering basic, you know, questions about your business and the growth of your company for so many organizations that we work with. So it's, I think one of the things that gets Viraj and I, and the rest of our company up every single morning, is knowing how important the work that we do for all of those use cases across industries, across company sizes. And it's really quite energizing. >> It was such a big focus this year at AWS re:Invent, the role of data. And I think one of the things that's exciting about the open AI and all the movement towards large language models, is that you can integrate data into these models, right? From outside, right? So you're starting to see the integration easier to deal with, still a lot of plumbing issues. So a lot of things happening. So I have to ask you guys, what is the state of the data orchestration area? Is it ready for disruption? Is it already been disrupted? Would you categorize it as a new first inning kind of opportunity or what's the state of the data orchestration area right now? Both, you know, technically and from a business model standpoint, how would you guys describe that state of the market? >> Yeah, I mean I think, I think in a lot of ways we're, in some ways I think we're categoric rating, you know, schedulers have been around for a long time. I recently did a presentation sort of on the evolution of going from, you know, something like KRON, which I think was built in like the 1970s out of Carnegie Mellon. And you know, that's a long time ago. That's 50 years ago. So it's sort of like the basic need to schedule and do something with your data on a schedule is not a new concept. But to our point earlier, I think everything that you need around your ecosystem, first of all, the number of data tools and developer tooling that has come out the industry has, you know, has some 5X over the last 10 years. And so obviously as that ecosystem grows and grows and grows and grows, the need for orchestration only increases. And I think, you know, as Astronomer, I think we, and there's, we work with so many different types of companies, companies that have been around for 50 years and companies that got started, you know, not even 12 months ago. And so I think for us, it's trying to always category create and adjust sort of what we sell and the value that we can provide for companies all across that journey. There are folks who are just getting started with orchestration and then there's folks who have such advanced use case 'cuz they're hitting sort of a ceiling and only want to go up from there. And so I think we as a company, care about both ends of that spectrum and certainly have want to build and continue building products for companies of all sorts, regardless of where they are on the maturity curve of data orchestration. >> That's a really good point Paola. And I think the other thing to really take into account is it's the companies themselves, but also individuals who have to do their jobs. You know, if you rewind the clock like five or 10 years ago, data engineers would be the ones responsible for orchestrating data through their org. But when we look at our customers today, it's not just data engineers anymore. There's data analysts who sit a lot closer to the business and the data scientists who want to automate things around their models. So this idea that orchestration is this new category is spot on, is right on the money. And what we're finding is it's spreading, the need for it, is spreading to all parts of the data team naturally where Airflows have emerged as an open source standard and we're hoping to take things to the next level. >> That's awesome. You know, we've been up saying that the data market's kind of like the SRE with servers, right? You're going to need one person to deal with a lot of data and that's data engineering and then you're going to have the practitioners, the democratization. Clearly that's coming in what you're seeing. So I got to ask, how do you guys fit in from a value proposition standpoint? What's the pitch that you have to customers or is it more inbound coming into you guys? Are you guys doing a lot of outreach, customer engagements? I'm sure they're getting a lot of great requirements from customers. What's the current value proposition? How do you guys engage? >> Yeah, I mean we've, there's so many, there's so many. Sorry Raj, you can jump in. - >> It's okay. So there's so many companies using Airflow, right? So our, the baseline is that the open source project that is Airflow that was, that came out of Airbnb, you know, over five years ago at this point, has grown exponentially in users and continues to grow. And so the folks that we sell to primarily are folks who are already committed to using Apache Airflow, need data orchestration in the organization and just want to do it better, want to do it more efficiently, want to do it without managing that infrastructure. And so our baseline proposition is for those organizations. Now to Raj's point, obviously I think our ambitions go beyond that, both in terms of the personas that we addressed and going beyond that data engineer, but really it's for, to start at the baseline. You know, as we continue to grow our company, it's really making sure that we're adding value to folks using Airflow and help them do so in a better way, in a larger way and a more efficient way. And that's really the crux of who we sell to. And so to answer your question on, we actually, we get a lot of inbound because they're are so many - >> A built-in audience. >> In the world that use it, that those are the folks who we talk to and come to our website and chat with us and get value from our content. I mean the power of the open source community is really just so, so big. And I think that's also one of the things that makes this job fun, so. >> And you guys are in a great position, Viraj, you can comment, to get your reaction. There's been a big successful business model to starting a company around these big projects for a lot of reasons. One is open source is continuing to be great, but there's also supply chain challenges in there. There's also, you know, we want to continue more innovation and more code and keeping it free and and flowing. And then there's the commercialization of product-izing it, operationalizing it. This is a huge new dynamic. I mean, in the past, you know, five or so years, 10 years, it's been happening all on CNCF from other areas like Apache, Linux Foundation, they're all implementing this. This is a huge opportunity for entrepreneurs to do this. >> Yeah, yeah. Open source is always going to be core to what we do because, you know, we wouldn't exist without the open source community around us. They are huge in numbers. Oftentimes they're nameless people who are working on making something better in a way that everybody benefits from it. But open source is really hard, especially if you're a company whose core competency is running a business, right? Maybe you're running e-commerce business or maybe you're running, I don't know, some sort of like any sort of business, especially if you're a company running a business, you don't really want to spend your time figuring out how to run open source software. You just want to use it, you want to use the best of it, you want to use the community around it. You want to take, you want to be able to google something and get answers for it. You want the benefits of open source. You don't want to have, you don't have the time or the resources to invest in becoming an expert in open source, right? And I think that dynamic is really what's given companies like us an ability to kind of form businesses around that, in the sense that we'll make it so people get the best of both worlds. You'll get this vast open ecosystem that you can build on top of, you can benefit from, that you can learn from, but you won't have to spend your time doing undifferentiated heavy lifting. You can do things that are just specific to your business. >> It's always been great to see that business model evolved. We used to debate 10 years ago, can there be another red hat? And we said, not really the same, but there'll be a lot of little ones that'll grow up to be big soon. Great stuff. Final question, can you guys share the history of the company, the milestones of the Astronomer's journey in data orchestration? >> Yeah, we could. So yeah, I mean, I think, so Raj and I have obviously been at astronomer along with our other founding team and leadership folks, for over five years now. And it's been such an incredible journey of learning, of hiring really amazing people. Solving again, mission critical problems for so many types of organizations. You know, we've had some funding that has allowed us to invest in the team that we have and in the software that we have. And that's been really phenomenal. And so that investment, I think, keeps us confident even despite these sort of macroeconomic conditions that we're finding ourselves in. And so honestly, the milestones for us are focusing on our product, focusing on our customers over the next year, focusing on that market for us, that we know can get value out of what we do. And making developers' lives better and growing the open source community, you know, and making sure that everything that we're doing makes it easier for folks to get started to contribute to the project and to feel a part of the community that we're cultivating here. >> You guys raised a little bit of money. How much have you guys raised? >> I forget what the total is, but it's in the ballpark of 200, over $200 million. So it feels good - >> A little bit of capital. Got a little bit of cash to work with there. Great success. I know it's a Series C financing, you guys been down, so you're up and running. What's next? What are you guys looking to do? What's the big horizon look like for you? And from a vision standpoint, more hiring, more product, what is some of the key things you're looking at doing? >> Yeah, it's really a little of all of the above, right? Like, kind of one of the best and worst things about working at earlier stage startups is there's always so much to do and you often have to just kind of figure out a way to get everything done, but really invest in our product over the next, at least the next, over the course of our company lifetime. And there's a lot of ways we wanting to just make it more accessible to users, easier to get started with, easier to use all kind of on all areas there. And really, we really want to do more for the community, right? Like I was saying, we wouldn't be anything without the large open source community around us. And we want to figure out ways to give back more in more creative ways, in more code driven ways and more kind of events and everything else that we can do to keep those folks galvanized and just keeping them happy using Airflow. >> Paola, any final words as we close out? >> No, I mean, I'm super excited. You know, I think we'll keep growing the team this year. We've got a couple of offices in the US which we're excited about, and a fully global team that will only continue to grow. So Viraj and I are both here in New York and we're excited to be engaging with our coworkers in person. Finally, after years of not doing so, we've got a bustling office in San Francisco as well. So growing those teams and continuing to hire all over the world and really focusing on our product and the open source community is where our heads are at this year, so. >> Congratulations. - >> Excited. 200 million in funding plus good runway. Put that money in the bank, squirrel it away. You know, it's good to kind of get some good interest on it, but still grow. Congratulations on all the work you guys do. We appreciate you and the open sourced community does and good luck with the venture. Continue to be successful and we'll see you at the Startup Showcase. >> Thank you. - >> Yeah, thanks so much, John. Appreciate it. - >> It's theCube conversation, featuring astronomer.io, that's the website. Astronomer is doing well. Multiple rounds of funding, over 200 million in funding. Open source continues to lead the way in innovation. Great business model. Good solution for the next gen, Cloud, scale, data operations, data stacks that are emerging. I'm John Furrier, your host. Thanks for watching. (soft music)

>>Welcome back to the Cube, as a continued coverage here from AWS Reinvent 22. It's day three of our coverage here at the Venetian in Las Vegas, and we're part of the AWS Global Startup Showcase. With me to talk about what Kong's to in that regard is Marco Palladino, who's the, the CTO and the co-founder of Con Marco. Good >>To see you. Well, thanks for having me >>Here. Yeah, I was gonna say, by the way, I, I, you've got a beautiful exhibit down on the show floor. How's the week been for you so far as an exhibitor here? >>It's been very busy. You know, to this year we made a big investment at the WS reinvent. You know, I think this is one of the best conferences in the industry. There is technology developers, but it's also business oriented. So you can learn about all the business outcomes that our, you know, customers or, you know, people are trying to make when, when adopting these new technologies. So it's very good so far. >>Good, good, good to hear. Alright, so in your world, the API world, you know, it used to be we had this, you know, giant elephant. Now we're cutting down the little pieces, right? That's right. We're all going micro now these days. That's right. Talk about that trend a little bit, what you're seeing, and we'll jump in a little deeper as to how you're addressing that. >>Well, I think the industry learned a long time ago that running large code bases is actually quite problematic when it comes to scaling the organization and capturing new opportunities. And so, you know, we're transitioning to microservices because we want to get more opportunities in our business. We want to be able to create new products, fasters, we want to be able to leverage existing services or data that we have built, like an assembly line of software, you know, picking up APIs that other developers are building, and then assemble them together to create new experiences or new products, enter new markets. And so microservices are fantastic for that, except microservices. They also introduce significant concerns on the networking layer, on the API layer. And so this is where Kong specializes by providing API infrastructure to our customers. >>Right. So more about the problems, more about the challenges there, because you're right, it, opportunities always create, you know, big upside and, and I, I don't wanna say downside, but they do introduce new complexities. >>That's right. And introducing new complexity. It's a little bit the biggest enemy of any large organization, right? We want to reduce complexity, we want to move faster, we want to be more agile, and, and we need an API vision to be able to do that. Our teams, you know, I'm speaking with customers here at Reinvent, they're telling me that in the next five years, the organization is going to be creating more APIs than all the APIs they've created up until now. Right? So how do you >>Support, that's a mind boggling number, right? >>It's mind boggling. Yeah, exactly. How do you support that type of growth? And things have been moving so fast. I feel like there is a big dilemma in, you know, with certain organizations where, you know, we have not taught a long term strategy for APIs, whereas we do have a long term strategy for our business, but APIs are running the business. We must have a long term strategy for our APIs, otherwise we're not gonna be able to execute. And that's a big dilemma right now. Yeah. >>So, so how do we get the horse back in front of the cart then? Because it's like you said, it's almost as if we've, we're, we're reprioritizing, you know, incorrectly or inaccurately, right? You're, you're getting a little bit ahead of ourselves. >>Well, so, you know, whenever we have a long-term strategy for pretty much anything in the organization, right? We know what we want to do. We know the outcome that we want to achieve. We work backwards to, you know, determine what are the steps that are gonna bring us there. And, and the responsibility for thinking long term in, in every organization, including for APIs at the end of the day, always falls on the leaders and the should on the shoulders of the leadership and, and to see executives of the organization, right? And so we're seeing, you know, look at aws by the way. Look at Amazon. This conference would not have been possible without a very strong API vision from Amazon. And the CEO himself, Jeff Bezos, everybody talks about wanting to become an API first organization. And Amazon did that with the famous Jeff Bezos mandate today, aws, it's a hundred billion revenue for Amazon. You see, Amazon was not the first organization with, with an e-commerce, but if it was the first one that married a very strong e-commerce business execution with a very strong API vision, and here we are. >>So yeah, here we are putting you squarely in, in, in a pretty good position, right? In terms of what you're offering to the marketplace who has this high demand, you see this trend starting to explode. The hockey sticks headed up a little bit, right? You know, how are you answering that call specifically at how, how are you looking at your client's needs and, and trying to address what they need and when they need it, and how they need it. Because everybody's in a kind of a different place right now. >>Right? That's exactly right. And so you have multiple teams at different stages of their journey, right? With technology, some of them are still working on legacy, some of them are moving to the cloud. Yep. Some of them are working in containers and in microservices and Kubernetes. And so how do you, how do we provide an API vision that can fulfill the needs of the entire organization in such a way that we reduce that type of fragmentation and we don't introduce too much complexity? Well, so at con, we do it by essentially splitting the API platform in three different components. Okay. One is API management. When, whenever we want to expose APIs internally or to an ecosystem of partners, right? Or to mobile, DRA is a service mesh. You know, as we're splitting these microservices into smaller parts, we have a lot of connectivity, all, you know, across all the services that the teams are building that we need to, to manage. >>You know, the network is unreliable. It's by default, not secure, not observable. There is nothing that that works in there. And so how do we make that network reliable without asking our teams to go and build these cross-cut concerns whenever they create a new service. And so we need a service match for that, right? And then finally, we could have the best AP infrastructure in the world, millions of APIs and millions of microservices. Everything is working great. And with no API consumption, all of that would be useless. The value of our APIs and the value of our infrastructure is being driven by the consumption that we're able to drive to all of these APIs. And so there is a whole area of API productivity and discovery and design and testing and mocking that enables the application teams to be successful with APIs, even when they do have a, the proper API infrastructure in place that's made of meshes and management products and so on and so forth. Right. >>Can you gimme some examples? I mean, at least with people that you've been working with in terms of addressing maybe unique needs. Cuz again, as you've addressed, journeys are in different stages now. Some people are on level one, some people are on level five. So maybe just a couple of examples Yeah. Of clients with whom you've been working. Yeah, >>So listen, I I was talking with many organizations here at AWS Reinvent that are of course trying to migrate to the cloud. That's a very common common transformation that pretty much everybody's doing in the world. And, and how do you transition to the cloud by de-risking the migration while at the same time being able to get all the benefits of, of running in the cloud? Well, we think that, you know, we can do that in two, two ways. One, by containerizing our workloads so that we can make them portable. But then we also need to lift and shift the API connectivity in such a way that we can determine how much traffic goes to the legacy and how much traffic goes to the new cloud infrastructure. And by doing that, we're able to deal with some of these transformations that can be quite complex. And then finally, API infrastructure must support every team in the organization. >>And so being able to run on a single cloud, multi-cloud, single cluster, multi cluster VMs containers, that's important and essential because we want the entire organization to be on board. Because whenever we do not do that, then the developers will make short term decisions that are not going to be fitting into the organizational outcomes that we want to achieve. And we look at any outcome that your organization wants to achieve the cloud transformation, improving customer retention, creating new products, being more agile. At the end of the day, there is an API that's powering that outcome. >>Right? Right. Well, and, and there's always a security component, right? That you have to be concerned about. So how are you raising that specter with your clients to make them aware? Because sometimes it, I wouldn't say it's an afterthought, but sometimes it's not the first thought. And, and obviously with APIs and with their integral place, you know, in, in the system now security's gotta be included in that, right? >>API security is perhaps the biggest, biggest request that we're hearing from customers. You know, 83% of the world's internet traffic at the end of the day runs on APIs, right? That's a lot of traffic. As a matter of fact, APIs are the first attack vector for any, you know, malicious store party. Whenever there is a breach, APIs must be secured. And we can secure APIs on different layers of our infrastructure. We can secure APIs at the L four mesh layer by implementing zero trust security, for example, encrypting all the traffic, assigning an identity to every service, removing the concept of trust from our systems because trust is exploitable, right? And so we need to remove the cut zero trust, remove the concept of trust, and then once we have that underlying networking that's being secure and encrypted, we want to secure access to our APIs. >>And so this is the typical authentication, authorization concerns. You know, we can use patterns like op, op or opa open policy agent to create a security layer that does not rely on the team's writing code every time they're creating a new service. But the infrastructure is enforcing the type of layer. So for example, last week I was in Sweden, as a matter of fact speaking with the largest bank in Sweden while our customers, and they were telling us that they are implementing GDPR validation in the service mesh on the OPPA layer across every service that anybody's building. Why? Well, because you can embed the GDPR settings of the consumer into a claim in a gel token, and then you can use OPPA to validate in a blanket way that Jo Token across every service in the mesh, developers don't have to do that. It just comes out of the box like that. And then finally, so networking, security, API security for access and, and management of those APIs. And then finally we have deep inspection of our API traffic. And here you will see more exotic solutions for API security, where we essentially take a subset of our API traffic and we try to inspect it to see if there is anybody doing anything that they shouldn't be doing and, and perhaps block them or, you know, raise, raise, raise the flag, so to speak. >>Well, the answer is probably yes, they are. Somebody's trying to, somebody's trying to, yeah, you're trying to block 'em out. Before I let you go, you've had some announcements leading up here to the show that's just to hit a few of those highlights, if you would. >>Well, you know, Kong is an organization that you know, is very proud of the technology that we create. Of course, we started with a, with the API gateway Con Gateway, which was our first product, the most adopted gateway in the world. But then we've expanded our platform with service mesh. We just announced D B P F support in the service mesh. For example, we made our con gateway, which was already one of the fastest gateway, if not the fastest gateway out there, 30% faster with Con Gateway 3.0. We have shipped an official con operator for Kubernetes, both community and enterprise. And then finally we're doubling down on insomnia, insomnia's, our API productivity application that essentially connects the developers with the APIs that are creating and allows them to create a discovery mechanism for testing, mocking the bagging, those APIs, all of this, we of course ship it OnPrem, but then also on the cloud. And you know, in a cloud conference right now, of course, cloud, right? Right. Is a very important part of our corporate strategy. And our customers are asking us that. Why? Because they don't wanna manage the software, they want the API platform, they don't, don't wanna manage it. >>Well, no, nobody does. And there are a few stragglers, >>A few, a few. And for them there is the on-prem >>Platform. Fine, let 'em go. Right? Exactly. But if you wanna make it a little quick and dirty, hand it off, right? Oh, >>That's exactly right. Yes. >>Let Con do the heavy lifting for you. Hey Marco, thanks for the time. Yeah, thank you so much. We appreciate, and again, congratulations on what appears to be a pretty good show for you guys. Yeah, thank you. Well done. All right, we continue our discussions here at aws. Reinvent 22. You're watching the Cube, the leader in high tech coverage. >>Okay.

>>Oh, welcome back to the Cube here at aws Reinvent 22. As we continue our coverage here, the AWS Global Showcase, the Startup Showcase, John Wall is here hosting for the Cube as we've been here all week. Hope you're enjoying our coverage here. This is day three, by the way. We're wrapping it up shortly with us to talk about what's going on in the, kind of the hotel world in it and what's going on in the cloud, especially at I hg is Eric Norman, head of infrastructure, architecture, and innovation at I H G Hotels and Resorts. Eric, good to see you, >>Sir. Oh, thank you. And thank you for inviting me. Yeah, >>You bet. Glad to have you board here on the queue. First time, I think too, by the way, right? >>It is. And can I just tell you who IHG is >>Real quick? Yeah, wait a second. First I want another rest. I got Introduc to Rod Stuller, who is the Vice president and of Solutions marketing at Aviatrix and Rod. Good to see you, sir. Thanks a lot. Now let's talk about I ih. >>Great. Well, IHGs a a hospitality company, it's been around for 200 years, that has 17 brands globally in over a hundred countries. We sleek, you know, up could up to 888,000 people a night. So it's a pretty large company that we compete with, you know, all the hotel companies globally. >>So let's talk about your, your footprint right now in, in terms of what your needs are, because you've mentioned obviously a lot of, you have a lot of customers needs, you have a lot of internal stakeholder needs. Yeah. So just from that perspective, how are you balancing out, you know, the products you wanna launch as opposed to the, on the development side and the maintenance side? >>Yeah, I mean we, we have focused our, our attention to our, our guests and our hotels globally and, and taking technology and from a foundation, getting it at, at the edge so that way the consumer and the hotel owner can deliver a quality product to a guest experience. You know, we've have moved larger, a large deployment of our mission critical applications over the last five years really, of moving into more SaaS and infrastructure like AWS and GCP and, and leveraging their global scale to be able to deliver at the edge or get closer to the edge. And so we've, you know, I'm pretty sure you've seen, you know, kind of people building, you know, mission critical apps. You know, probably in the last three years it's probably escalating and more of like a hockey stick of moving stuff. I'd love to hear what AVIA is seeing. Oh >>Yeah. Now we're, we're seeing that quite a bit, right? As people move into the cloud, it's now business critical applications that are going there. So good enough isn't good enough anymore, right? It has to be, you know, a powerful capability that's business critical, can support that, give people the ability to troubleshoot it when something goes wrong. And then multi-cloud, you mentioned a couple different cloud companies, a lot of enterprises are moving to multiple clouds and you don't want to have to do it differently in every cloud. You want a infrastructure management layer that allows you to do that across >>Clouds. So how do you go about that, you know, deciding what goes where. I mean, it sounds like a simple question, but, but if you are dealing in a lot of different kinds of environments, different needs and different requirements, whatever, you know, how are you sorting out, delegating, you know, you know, you're, you're you're gonna be working here, you're gonna be >>Working there. Yeah. So we built some standards base that says, you know, certain types of apps, you know, transactional base, you know, go to this cloud provider and data analytics that's gonna go to another, another cloud provider based on our decision of key capability, native capability, and, and also coverage. You know, cuz we are in China, right? You know, you know, I, I've gotta be able to get into China and, and build not only a network that can support that, but also business apps locally to meet, compete with compliance, regulatory type activities. I mean, even in, in the US market, I got, you know, California privacy laws, you know, you have globally, you've gotta deal with getting data applications into compliance for those globally, right? >>Yeah. So, so you got that compliance slash governance Yeah. Issue. Huge issue. Yeah. I would think for you, you gotta decide who's gonna get to what when, and also we have to meet certain regulatory standards as you pointed out. And not just there, but you got European footprint, right? I mean, you're global. Yeah. So, so you know, handling that kind of scope or scale, what kind of nightmares or challenges does that provide you and how's Aviatrix helping you solve >>That? Yeah, in the early days, you know, we were using cloud native, you know, constructs for networking and a little bit of a security type angle to it. What we found was, you know, you can't get the automation you need. You can't get the, the scalability, you know, cuz we're, we're trying to shift left our, you know, our DevOps and our ability to deploy infrastructure. Aviatrix had come in and, and provided a, a solution that gets us there quicker than anybody else. It's allow us to, you know, build a mesh network across all our regions globally. I'm able to deploy, you know, new landing zones or, you know, public cloud fairly quickly with my, you know, networking construct. We also, we found that because we are a multi hybrid cloud, we, we introduced on the edge a a new network. We had to introduce a performance hub architecture that's using Equinix that sits in every region in every public cloud and partner. Cuz all our partners, you know, we, we've moved a lot of stuff to sas. You know, Amadeus is our centralized reservation system. That's our key, you know? Sure. You know, reservation tool, it's so sourced out. I need to bring them in and I need to get data that's closer to where, in a region to where it needs the land so I can process it. Right. >>And it's a big world out there too. I mean, you're, you're not in your head Rod. So talk about if you would share some of the, the aviatrix experience in that regard. When you have a client like this that has these, you know, multinational locations and, and yet you're looking for some consistency and some uniformity. You don't, you know, you can't be reinventing the wheel every time something pops up, right? >>Right. No. And then, and it's about agility and speed and, you know, being able to do it with less people than you used to have to do things, right? You, you want to be able to give the developers what they need when they need it. There was a time when people were going around it, swiping their credit card and, and saying, it doesn't give me what I need. And so cloud is supposed to change that. So we're trying to deliver the ability to do that for the developers a lot faster than had been done in the past. But at the same time, giving the enterprise the controls, the security, the compliance that they need. And sometimes those things got in the way, but now we're building systems that allow that to happen at, at the piece that developers needed to happen. >>But what Rod said about, you know, one of the big things you sparked my thinking is it also, you know, building a overlay of the cloud native construct allows for visibility that, you know, you didn't have, you know, from a developer or even a operations day two operations, now you get that visibility into the network space and controls and management of that space a lot easier now, you know? >>Yeah. I mean, business critical applications, right? People, the people, the business does not care about networking, right? They see it as electricity and if it's down somebody else's problem to fix it. But the people who do need to keep it up, they need the telemetry. They need the ability to understand, are we trending in the wrong direction? Should we be doing something so that we don't get to the point where it goes down? And that's the kind of information that we're providing in this multi-cloud environment. You mentioned Equinix, we, we just have a partnership with Equinix where we're extending the cloud operational model that Aviatrix delivers all the way out to Equinix and that global fabric that you're talking about. So this is allowing the, the comp companies to have that visibility, that operational ability all the way globally. >>Yeah. Because you know, when you start building all these clouds now and multi regions, multiple AZs or different cloud providers or SaaS providers, you're moving data all over the place. And if you, if you don't have a single pane of glass to see that entire network and be able to route stuff accordingly, it's gonna be a zoo. It's not gonna >>Work. We were, I was talking earlier with, with another guest and we were just talking about companies in your case, I, I IHG kind of knowing what you have and it's not like such a basic thing he said, but yeah, you'd be surprised how many people don't know what they have. Oh, yeah. And so they're trying to provide that visibility and, and, and awareness. So, so I'm kind of curious because you were just the next interview up, so sorry Ken, but, but do you know what you have, I mean, are you learning what you have or is how do you identify, prioritize? How valuable is this asset as opposed to this can wait? I mean, is that still an ongoing process for >>You? It, it's definitely an ongoing process. I mean, we've done over the last three years of constantly assessing all our inventory of what we have, making sure we have the right mo roadmaps for each of the apps and products that we have. Cause we've turned to more of a product driven organization and a DevOps and we're, we're moving more and more product teams onto that DevOps process. Yep. So we can shift left a lot of the activities that developer in the past had to go over a fence to ask for help and, and, you know, kind of the automation of the network and the security built in allows us to be able to shift that left. >>Did that, I, you were saying too three years, right? You've been on, on this path Yep. Going back then to 2019 right. Pandemic hits, right. The world changes. How has that affected this three year period for you? And where are you in terms of where you expected to be and, and Yep. And then what's your, what are your headlights seeing down the road as to what your, your eventual journey, how you want that to end? >>I probably, the biggest story that we have a success story is when the pandemic did happen, you know, all our call centers, all agents had to go home. We were able within 30 days be able to bring up remote desktops, you know, workspaces an a uws and give access to globally in China and in Singapore and in the Americas. There's >>No small task there, >>That's for sure. So we built a desktop, certified it, and, and agents were able to answer calls for guests, you know, you know, so it was a huge success to us. Sure. It did slow down. I mean, during the pandemic it did slow us down from what gets migrated. You know, our focus is, you know, again, back to what I was saying earlier is around our guests and our loyalty and, you know, how do we give value back to our hotel owners and our guests? >>And how do you measure that? I mean, how do you know that what you're doing is working with, with that key audience? >>We'd measured by, you know, one occupa >>There so many, how many people do we have in the rooms? Right? But in terms of the interface, in terms of the effectiveness, the applications, in terms of what you're offering. Yeah. >>It gets back to uptime of our systems and you know, being able to deploy an application in multiple regions elevates the availability of the product to our guest. You know, the longer I'm up, the more revenue I can produce. Right. So, you know, so we, we try to, you know, we measure also guest satisfaction at the properties, you know, them using our tech and that kind of stuff to >>Be so you surveying just to find out what, how they feel about, so some, >>Cause we have a lot of tech inside of our hotels that allow for, we have ISG connect, which allows for people to go from one hotel another and not ask for passwords and, you know, that kind of stuff. >>That would not be made by the way. I'd be begging for help. Let's talk about skills, because I hear that a lot. Talk a lot about that this week. Hearing that, that, you know, the advancement of knowledge is obviously a very powerful thing, but it's also a bit of a shortcoming right now in terms of, of having a need for skills and not having that kind of firepower horsepower on your bench. What, what do you see in that regard? And, and first off, what did you see about it? And then I'll follow >>Up with Yeah, I mean, over our journey, it started off where you didn't have the skills, you know, you didn't have the skill from an operations engineering architecture. So we went on a, you know, you know, how do we build training programs? How do we get, you know, tools to, to either virtual training, bringing teachers, we built, you know, daily, our weekly calls where we bring our experts from our vendors in there to be able to ask questions to help engineering people or architecture people or operations to ask questions and get answers. You know, we, we've been on a role of, you know, upscaling over the last three years and we continue to drive that, you know, we have lunch and learns that we bring people to. Yep. You know, and, and we, and we, we ta tailor the, the content for that training based on what we are consuming and what we're using as opposed to just a, you know, a broad stroke of, of public cloud or, it's >>Almost like you don't have to be holistic about it. You just need to, what do you need to know to >>Make >>Them successful, to be better at what you're doing here? Right. Sure. >>And that's been huge. And, >>And yeah, we, and we have a program called ace, which is AVIATRIX certified engineer. And there's a bunch of different types of classes. So if you're a networking person in the past it's like A C C I E, but we have about 18,000 people over the last three years who have gone through that training. One of them. One of them, right? Is that right? Yeah. Yeah. And, and this is not necessarily about aviatrix. What we're doing is trying to give multi-cloud, you know, networking expertise because a lot of the people that we're talking about are coming from the data center world. And networking is so different in the cloud. We're helping them understand it's not as scary as they might think. Right. If your whole career has been networking in the data center and all of a sudden there's this cloud thing that you don't really understand, you need somebody to help you sort of get there. And we're doing that in a multi-cloud way. And we have all kinds of different levels to teach people how to do, do infrastructure as code. That's another thing, you know, data center guys, they never did infrastructure as code. It was, you had to bolt it in and plug stuff in. Right. But now things are being done much faster with infrastructure as code. And we're teaching people how >>To do that. Yeah. I mean, yesterday, one of the keynotes is about the partner in the, the marketplace. And they use the image imagery of, of marathon runner, you know, a marathon runner. Yeah. You could do a marathon by yourself, but if you want to improve and become a, a great marathon runner, you need a coach, you need nutritionist, you need people running with you to, to make that engine go faster a little bit. Yeah, exactly. And you know, having a partner like Aviatrix helps you know the team to be successful. >>Well, it is, it is a marathon, not a sprint. That's for sure. And you've been on this kind of three year jog. You might feel like you've been running a marathon a little bit, but it sounds like you're really off to a great start and, and have a pretty good partnership here. So thank you. Congratulations on that, Eric. Thank you for being with us. And Rod, same to you. Thank you. Appreciate the time here on the AWS Global Showcase. I'm John Wal, you're watching The Cube. We're out in Las Vegas and of course the cube, as you well know, is the leader in high tech coverage.

well hello everybody John Wallace here on thecube he's continuing our segments here on the AWS Global startup showcase we are at day three of Reinventing irking Zhang is joining us now he is the CEO co-founder of Jupiter one um first off before we get going talking about you know security and big world for you guys I know what's your take on the show what's been going on out here at re invent yeah yeah ring event has been one of my favorite shows there's a lot of people here there's a lot of topics of course it's not just cyber security a lot of cloud infrastructure and just technology in general so you get a lot you know if you go walk the floor you see a lot of vendors you look at us go into sessions you can learn a lot but you're the Hot Topic right everybody's focused on Cyber yeah big time and with good reason right because as we know the Bad actors are getting even smarter and even faster and even more Nimble so just paint the landscape for me here in general right now as you see uh security Cloud Security in particular and and kind of where we are in that battle well we are clearly not winning so I think that in itself is a bit of a uh interesting problem right so as a it's not just Cloud security if you think about cyber security in general as an industry it has it has not been around for that long right but if you just look at the history of it uh we haven't done that while so uh pick another industry say medicine which has been around forever and if you look at the history of Medicine well I would argue you has done tremendously well because people live longer right when you get sick you get access to health care and yeah exactly you have Solutions and and you can see the trend even though there are problems in healthcare of course right but the trend is is good it's going well but not in cyber security more breaches more attacks more attackers we don't know what the hell we're doing with that many solutions and you know that's been one of my struggles as a former CSO and security practitioner for many years you know why is it that we're not getting better all right so I'm going to ask you the question yeah okay why aren't we getting better you know how come we can't stay ahead of the curve on this thing that for some reason it's like whack-a-mole times a hundred every time we think we solve one problem we have a hundred more that show up over here exactly and we have to address that and and our attention keeps floating around yeah I think you said it right so because we're taking this guacamole approach and we're looking for the painkiller of the day and you know we're looking for uh the Band-Aids right so and then we ended up well I I think to be fair to be fair to your industry the industry moves so quickly technology in general moves so quickly and security has been playing catch-up over time we're still playing catch-up so when you're playing catch-up you you can almost only uh look at you know what's the painkiller of what's the band name of the day so I can stop the bleeding right but I do think that we're we're to a point or we have enough painkillers and Band-Aids and and we need to start looking at how can we do better fundamentally with the basics and do the basics well because a lot of times the basics that get you into trouble so fundamentally the foundation I if I hear you right what you're saying is um you know quick changing industry right things are moving rapidly but we're not blocking and tackling we're not doing the X's and O's and so forget changing and we we got to get back to the basis and do those things right exactly you can only seem so simple it seems so simple but it's so hard right so you can you can think about you know uh even in case of building a starter building a company and and in order at one point right so we're blocking uh blocking tackling and then when we grow to a certain size we have to scale we have to figure out how to scale the business this is the same problem that happens in security as an industry we've been blocking happening for so long you know we're the industry is so young but we're to a point that we got to figure out how to scale this scale this in a fundamentally different way and I'll give you some example right so so what when we say the basics now it's easy to to think that say users should have MFA enabled is one of the basics right or another Basics will be you have endpoint protection on your devices you know maybe it's Cloud strike or Sentinel one or carbon black or whatever but the question being how do you know it is working 100 of the time right how do you know that how do you know right you find out too exactly that's right and how do you know that you have 100 coverage on your endpoints those Solutions are not going to tell you because they don't know what they don't know right if it's not enabled if it's not you know what what's the negative that you are not seeing so that's one of the things that you know that's in the basic state that you're now covering so the fundamentals it really goes to these five questions that I think that nobody has a really good answer for until now so the five questions goes what do I have right is it important what's important out of all the things I have you have a lot right you could have millions of things what important now for those that are important does it have a problem and if it has a problem who can fix it because the reality is in most cases security teams are not the ones fixing the problems they're they're the ones identical they're very good at recognizing but not so good exactly identifying the owner who can fix it right right could be could be business owner could be Engineers so the the asset ownership identification right so so these four questions and and then over time you know whether it's over a week or a month or a quarter or a year am I getting better right and then you just keep asking these questions in different areas in different domains with a different lens right so maybe that's endpoints maybe that's Cloud maybe that's you know users maybe that's a product and applications right but it really boils down to these five questions that's the foundation for any good security program if you can do that well I think we cover a lot of bases and we're going to be in much better shape than we have been all right so where do you come in man Jupiter one in terms of what you're providing because obviously you've identified this kind of pyramid yes this hierarchy of addressing needs and I assume obviously knowing you as I do and knowing the company as I do you've got Solutions that's exactly right right and and we precisely answer those five questions right for uh any organization uh from a asset perspective right because all the the answers to all those these five questions are based in assets it starts with knowing what I have right right so the the overall challenge of cyber security being broke broken I I believe is fundamentally that people do not understand and cannot uh probably deal with the complexity that we have within our own environments so again like you know using uh medicine as an example right so in order to come up with the right medicine for either it's a vaccine for covid-19 or whether it is a treatment for cancer or whatever that case may be you have to start with the foundations of understanding both the pathogen and to the human body like DNA sequencing right without those you cannot effectively produce the right medicine in modern uh you know Medicine sure right so that is the same thing that's happening in cyber security you know we spend a lot of times you know putting band days in patches right and then we spend a lot of time doing attacker research from the outside but we don't fundamentally understand in a complete way what's the complexity within our own environment in terms of digital assets and that's that's almost like the DNA of your own work what is that kind of mind-blowing in a way that if again hearing you what you're talking about is saying that the first step is to identify what you have that's right so it seems just so basic that that I should know what I what's under my hood I should know what is valuable and what is not I should prioritize what I really need to protect and what maybe can go on the second shelf yeah it has been a tough problem since the beginning of I.T not just the beginning of cyber security right so in the history of I.T we have this thing called cmdb configuration management database it is supposed to capture the configurations of it assets now over time that has become a lot more complex and and there's a lot more than just it asset that we have to understand from a security and attack service perspective right so we have to understand I.T environments we have to understand Cloud environments and applications and users and access and data and as and all of those things then then we have to take a different approach of sort of a modern cmdb right so what is the way that we can understand all of those complexity within all of those assets but not just independently within those silos but rather in a connected way so we can not only understand the attack surface but only but also understand the attack path that connect the dots from one thing to another right because everything in the organization is actually connected if if there's any one thing that sits on an island right so if you say you have a a a a server or a device or a user that is on an island that is not connected to the rest of the organization then why have it right and it doesn't matter so it's the understanding of that connect connected tissue this entire map where this you know DNA sequencing equivalent of a digital organization is what Jupiter one provides right so that visibility of the fundamental you know very granular uh level of assets and resources to answer those five questions and how does that how do I get better at that then I mean I have you to help me but but internally within our organization um I mean I don't want to be rude but I mean do I have do I have the skill for that do I have um do I have the the internal horsepower for that or or is there some need to close that Gap and how do I do it you know I'll tell you two things right so so one you mentioned the worst skills right so let me start there so because this one is very interesting we also have a huge skills shortage in cyber security we will we've all heard that for years and and and and for a long time but if you dig deeper into it why is that why is that and you know we have a lot of you know talented people right so why do we still have a skills shortage now what's interesting is if you think about what we're asking security people to do is mind-boggling so if you if you get a security analyst to say hey I want to understand how to protect something or or how to deal with an incident and what you're asking the person to do is not only to understand the security concept and be a domain expert in security you're also asking the person to and understand at the same time AWS or other clouds or endpoints or code or applications so that you can properly do the analysis and the in the response it's it's impossible it's like you know if you have you have to have a person who's an expert in everything know everything about everything that's right it's impossible so so so that's that's one thing that we have to to resolve is how do we use technology like Jupiter one to provide an abstraction so that there's Automation in place to help the security teams be better at their jobs without having to be an expert in deep technology right just add the abstract level of understanding because you know we can we can model the data and and provide the analysis and visual visualization out of the box for them so they can focus on just the security practices so that's one and the second thing is we have to change the mindset like take vulnerability management as an example right so the mindset for vulnerability management has been how do I manage findings now we have to change it to the concept of more proactive and how to manage assets so let's think about uh you know say log4j right that that happened and uh you know when it happened everybody scrambles and said hey which which devices or which you know uh systems have log4j and you know it doesn't matter what's the impact we can fix it right going back to those questions that that I mentioned before right and then um and then they try to look for a solution at a time say well where's that silver bullet that can give me the answers now what what what we struggle with though is that you know I want to maybe ask the question where were you six months ago where were you six months ago where you could have done the due diligence and put something in place that help you understand all of these assets and connections so you can go to one place and just ask for that question when something like that you know hit the fan so so if we do not fundamentally change the mindset to say I have to look at things not from a reactive findings perspective but really starting from an asset-centric you know day one perspective to look at that and have this Foundation have this map build we can't get there right so it's like you know if I need direction I go to Google Maps right but the the reason that it works is because somebody has done the work of creating the map right right if you haven't if you don't have the map and you just at you know when the time you say I gotta go somewhere and you expect the map to magically happen to show you the direction it's not going to work right right I imagine there are a lot of people out there right now are listening to thinking oh boy you know and that's what Jupiter one's all about they're there to answer your oh boy thanks for the time of course I appreciate the insights as well it's nice to know that uh at least somebody is reminding us to keep the front door locked too that's just the back door the side doors keep that front door and that garage locked up too definitely um all right we'll continue our coverage here at AWS re invent 22 this is part of the AWS Global startup showcase and you're watching the cube the leader in high-tech coverage foreign

>>Well, hello everybody. John Walls here and continuing our coverage here at AWS Reinvent 22 on the queue. We continue our segments here in the Global Startup program, which of course is sponsored by AWS Startup Showcase, and with us to talk about any scale as the co-founder and CEO of the company, Robert and n, you are Robert. Good to see you. Thanks for joining us. >>Yeah, great. And thank you. >>You bet. Yeah. Glad to have you aboard here. So let's talk about Annie Scale, first off, for those at home and might not be familiar with what you do. Yeah. Because you've only been around for a short period of time, you're telling me >>Company's about >>Three years now. Three >>Years old, >>Yeah. Yeah. So tell us all about it. Yeah, >>Absolutely. So one of the biggest things happening in computing right now is the proliferation of ai. AI is just spreading throughout every industry has the potential to transform every industry. But the thing about doing AI is that it's incredibly computationally intensive. So if you wanna do do ai, you're not, you're probably not just doing it on your laptop, you're doing it across many machines, many gpu, many compute resources, and that's incredibly hard to do. It requires a lot of software engineering expertise, a lot of infrastructure expertise, a lot of cloud computing expertise to build the software infrastructure and distributed systems to really scale AI across all of the, across the cloud. And to do it in a way where you're really getting value out of ai. And so that is the, the problem statement that AI has tremendous potential. It's incredibly hard to do because of the, the scale required. >>And what we are building at any scale is really trying to make that easy. So trying to get to the point where, as a developer, if you know how to program on your laptop, then if you know how to program saying Python on your laptop, then that's enough, right? Then you can do ai, you can get value out of it, you can scale it, you can build the kinds of, you know, incredibly powerful applica AI applications that companies like Google and, and Facebook and others can build. But you don't have to learn about all of the distributed systems and infrastructure. It just, you know, we'll handle that for you. So that's, if we're successful, you know, that's what we're trying to achieve here. >>Yeah. What, what makes AI so hard to work with? I mean, you talk about the complexity. Yeah. A lot of moving parts. I mean, literally moving parts, but, but what is it in, in your mind that, that gets people's eyes spinning a little bit when they, they look at great potential. Yeah. But also they look at the downside of maybe having to work your way through Pike mere of sorts. >>So, so the potential is definitely there, but it's important to remember that a lot of AI initiatives fail. Like a lot of initiative AI initiatives, something like 80 or 90% don't make it out of, you know, the research or prototyping phase and inter production. Hmm. So, some of the things that are hard about AI and the reasons that AI initiatives can fail, one is the scale required, you know, moving. It's one thing to develop something on your laptop, it's another thing to run it across thousands of machines. So that's scale, right? Another is the transition from development and prototyping to production. Those are very different, have very different requirements. Absolutely. A lot of times it's different teams within a company. They have different tech stacks, different software they're using. You know, we hear companies say that when they move from develop, you know, once they prototype and develop a model, it could take six to 12 weeks to get that model in production. >>And that often involves rewriting a lot of code and handing it off to another team. So the transition from development to production is, is a big challenge. So the scale, the development to production handoff. And then lastly, a big challenge is around flexibility. So AI's a fast moving field, you see new developments, new algorithms, new models coming out all the time. And a lot of teams we work with, you know, they've, they've built infrastructure. They're using products out there to do ai, but they've found that it's sort of locking them into rigid workflows or specific tools, and they don't have the flexibility to adopt new algorithms or new strategies or approaches as they're being developed as they come out. And so they, but their developers want the flexibility to use the latest tools, the latest strategies. And so those are some of the main problems we see. It's really like, how do you scale scalability? How do you move easily from development and production and back? And how do you remain flexible? How do you adapt and, and use the best tools that are coming out? And so those are, yeah, just those are and often reasons that people start to use Ray, which is our open source project in any scale, which is our, our product. So tell >>Me about Ray, right? Yeah. Opensource project. I think you said you worked on it >>At Berkeley. That's right. Yeah. So before this company, I did a PhD in machine learning at Berkeley. And one of the challenges that we were running into ourselves, we were trying to do machine learning. We actually weren't infrastructure or distributed systems people, but we found ourselves in order to do machine learning, we found ourselves building all sorts of tools, ad hoc tools and systems to scale the machine learning, to be able to run it in a reasonable amount of time and to be able to leverage the compute that we needed. And it wasn't just us people all across, you know, machine learning researchers, machine learning practitioners were building their own tooling and infrastructure. And that was one of the things that we felt was really holding back progress. And so that's how we slowly and kind of gradually got into saying, Hey, we could build better tools here. >>We could build, we could try to make this easier to do so that all of these people don't have to build their own infrastructure. They can focus on the actual machine learning applications that they're trying to build. And so we started, Ray started this open source project for basically scaling Python applications and scaling machine learning applications. And, well, initially we were running around Berkeley trying to get all of our friends to try it out and, and adopt it and, you know, and give us feedback. And if it didn't work, we would debug it right away. And that slow, you know, that gradually turned into more companies starting to adopt it, bigger teams starting to adopt it, external contributors starting to, to contribute back to the open source project and make it better. And, you know, before you know it, we were hosting meetups, giving to talks, running tutorials, and the project was just taking off. And so that's a big part of what we continue to develop today at any scale, is like really fostering this open source community, growing the open source user base, making sure Ray is just the best way to scale Python applications and, and machine learning applications. >>So, so this was a graduate school project That's right. You say on, on your way to getting your doctorate and now you commercializing now, right? Yeah. I mean, so you're being able to offer it, first off, what a journey that was, right? I mean, who would've thought Absolutely. I guess you probably did think that at some point, but >>No, you know, when we started, when we were working on Ray, we actually didn't anticipate becoming a company, or we at least just weren't looking that far ahead. We were really excited about solving this problem of making distributed computing easy, you know, getting to the point where developers just don't have to learn about infrastructure and distributed systems, but get all the benefits. And of course, it wasn't until, you know, later on as we were graduating from Berkeley and we wanted to continue really taking this project further and, and really solving this problem that it, we realized it made sense to start a company. >>So help me out, like, like what, what, and I might have missed this, so I apologize if I did, but in terms of, of Ray's that building block and essential for your, your ML or AI work down the road, you know, what, what is it doing for me or what, what will it allow me to do in either one of those realms that I, I can't do now? >>Yeah. And so, so like why use Ray versus not using Ray? Yeah, I think the, the answer is that you, you know, if you're doing ai, you need to scale. It's becoming, if you don't find that to be the case today, you probably will tomorrow, you know, or the day after that. And so it's really increasingly, it's a requirement. It's not an option. And so if you're scaling, if you're trying to build these scalable applications you are building, you're either going to use Ray or, or something like Ray or you're going to build the infrastructure yourself and building the infrastructure yourself, that's a long journey. >>So why take that on, right? >>And many of the companies we work with don't want to be in the business of building and managing infrastructure. No. Because, you know, if they, they want their their best engineers to build their product, right? To, to get their product to market faster. >>I want, I want you to do that for me. >>Right? Exactly. And so, you know, we can really accelerate what these teams can do and, you know, and if we can make the infrastructure something they just don't have to think about, that's, that's why you would choose to use Ray. >>Okay. You know, between a and I and ml are, are they different animals in terms of what you're trying to get done or what Ray can do? >>Yeah, and actually I should say like, it's not just, you know, teams that are new teams that are starting out, that are using Ray, many companies that have built, already built their own infrastructure will then switch to using Ray. And to give you a few examples, like Uber runs all their deep learning on Ray, okay. And, you know, open ai, which is really at the frontier of training large models and, and you know, pushing the boundaries of, of ai, they train their largest models using Ray. You know, companies like Shopify rebuilt their entire machine learning platform using Ray, >>But they started somewhere else. >>They had, this is all, you know, like, it's not like the v1, you know, of their, of their machine learning infrastructure. This is like, they did it a different way before, this is like the second version or the third iteration of of, of how they're doing it. And they realize often it's because, you know, I mean in the case of, of Uber, just to give you one example, they built a system called hova for scaling deep learning on a bunch of GPUs. Right Now, as you scale deep learning on GPUs for them, the bottleneck shifted away from, you know, as you scale GPU's training, the bottleneck shifted away from training and to the data ingest and pre-processing. And they wanted to scale data ingest and pre-processing on CPUs. So now Hova, it's a deep learning framework. It doesn't do the data ingest and pre-processing on CPUs, but you can, if you run Hova on top of Ray, you can scale training on GPUs. >>And then Ray has another library called Ray Data you can, that lets you scale the ingest and pre-processing on CPUs. You can pipeline them together. And that allowed them to train larger models on more data before, just to take one example, ETA prediction, if you get in an Uber, it tells you what time you're supposed to arrive. Sure. That uses a deep learning model called d eta. And before they were able to train on about two weeks worth of data. Now, you know, using Ray and for scaling the data, ingestive pre-processing and training, they can train on much more data. You know, you can get more accurate ETA predictions. So that's just one example of the kind of benefit they were able to get. Right. Also, because it's running on top of, of Ray and Ray has this ecosystem of libraries, you know, they can also use Ray's hyper parameter tuning library to do hyper parameter tuning for their deep learning models. >>They can also use it for inference and you know, because these are all built on top of Ray, they inherit the like, elasticity and fault tolerance of running on top of Ray. So really it simplifies things on the infrastructure side cuz there's just, if you have Ray as common infrastructure for your machine learning workloads, there's just one system to, to kind of manage and operate. And if you are, it simplifies things for the end users like the developers because from their perspective, they're just writing a Python application. They don't have to learn how to use three different distributed systems and stitch them together and all of this. >>So aws, before I let you go, how do they come into play here for you? I mean, are you part of the showcase, a startup showcase? So obviously a major partner and major figure in the offering that you're presenting >>People? Yeah, well you can run. So any scale is a managed ray service. Like any scale is just the best way to run Ray and deploy Ray. And we run on top of aws. So many of our customers are, you know, using Ray through any scale on aws. And so we work very closely together and, and you know, we have, we have joint customers and basically, and you know, a lot of the value that any scale is adding on top of Ray is around the production story. So basically, you know, things like high availability, things like failure handling, retry alerting, persistence, reproducibility, these are a lot of the value, the values of, you know, the value that our platform adds on top of the open source project. A lot of stuff as well around collaboration, you know, imagine you are, you, something goes wrong with your application, your production job, you want to debug it, you can just share the URL with your, your coworker. They can click a button, reproduce the exact same thing, look at the same logs, you know, and, and, and figure out what's going on. And also a lot around, one thing that's, that's important for a lot of our customers is efficiency around cost. And so we >>Support every customer. >>Exactly. A lot of people are spending a lot of money on, on aws. Yeah. Right? And so any scale supports running out of the box on cheaper like spot instances, these preempt instances, which, you know, just reduce costs by quite a bit. And so things like that. >>Well, the company is any scale and you're on the show floor, right? So if you're having a chance to watch this during reinvent, go down and check 'em out. Robert Ashihara joining us here, the co-founder and ceo and Robert, thanks for being with us. Yeah. Here on the cube. Really enjoyed it. Me too. Thanks so much. Boy, three years graduate program and boom, here you are, you know, with off to the enterprise you go. Very nicely done. All right, we're gonna continue our coverage here on the Cube with more here from Las Vegas. We're the Venetian, we're AWS Reinvent 22 and you're watching the Cube, the leader in high tech coverage.