>> Announcer: Live from Las Vegas, it's theCUBE, covering AWS re:Invent 2019. Brought to you by Amazon Web Services and Intel, along with its ecosystem partners. >> Well, welcome back here on theCUBE. We're at AWS re:Invent 2019. And every once in a while, we have one of these fascinating interviews that really reaches beyond the technological prowess that's available today into almost the human fascination of work, and that's what we have here. >> Big story. >> Dave Vellante, John Walls. We're joined by Sebastien De Halleux, who is the CEO, oh, COO, rather, of a company called Saildrone, and what they feature is wind-powered flying robots, and they've undertaken a project called Seabed 2030 that will encompass mapping the world's oceans. 85% of the oceans, we know nothing about. >> That's right. >> And, yeah, they're going to combine this tremendous technology with 100 of these flying drones. So, Sebastien, we're really excited to have you here. Thanks for joining us, and wow, what a project! So, just paint the high-level view, I mean, not to have a pun here, but just to share with folks at home a little bit about the motivation of this and what gap you're going to fill. Then we'll get into the technology. >> So I think, you know, the first question is to realize the role of oceans and how they affect you on land and all of us. Half the air you breathe, half the oxygen you breathe, comes from the ocean. They cover 70% of the planet and drive global weather, they drive all the precipitation. They also drive sea-level rise, which affects coastal communities. They provide 20% of the protein, all the fish that we all eat. So, you know, it's a very, very important survival system for all of us on land. The problem is, it's also a very hostile environment, very dangerous, and so, we know very little about it. Because we study it with a few ships and buoys, but that's really a few hundred data points to cover 70% of the planet, whereas on land, we have billions of data points that are connected. So, that's why we're trying to fundamentally address, is deploying sensors in the ocean using autonomous surface vehicles, what we call Saildrones, which are essentially, think of them as autonomous sailboats, seven meters, 23 feet, long, bright orange thing with a five-meter-tall sail, which is harnessing wind power for propulsion and solar power for the onboard electronics. >> And then you've got sonar attached to that, that is what's going to do the-- >> The mapping itself. >> The underwater mapping, right, so you can look for marine life, you can look for geographical or topographical anomalies and whatever, and so, it's a multidimensional look using this sonar that, I think, is powered down to seven kilometers, right? >> That's right. >> So that's how far down, 20,000, 30,000 feet. >> That's right. >> So you're going to be able to derive information from it. >> You essentially describe it as, you're painting the ocean with sound. >> That's absolutely right, whereas if you wanted to take a picture of land, you could fly an airplane or satellite and take a photograph, light does not travel through water that well. And so, we use sound instead of light, but the same principle, which is that we send those pulses of sound down, and the echo we listen to from the seabed, or from fish or critters in the water column. And so, yes, we paint the ocean with sound, and then we use machine learning to transform this data into biomass, statistical biomass distribution, for example, or a 3-D surface of the seabed, after processing the sound data. >> And you have to discern between different objects, right? I mean, you (laughs) showed one picture of a seal sunbathing on one of these drones, right? Or is there a boat on the horizon? How do you do that? >> It's an extremely hard problem, because if a human is at sea looking through binoculars at things on the horizon, you're going to become seasick, right? So imagine the state of the algorithm trying to process this in a frame where every pixel is moving all the time, unlike on land, where you have at least a static frame of reference. So it's a very hard problem, and one of the first problems is training data. Where do you get all this training data? So our drones, hundreds of drones, take millions of pictures of the ocean, and then we train the algorithm using either labeled datasets or other source of data, and we teach them what is a boat on the horizon, what does that look like, and what's a bird, what's a seal. And then, in some hard cases, when you have a whale under the Saildrone or a seal lying on it, we have a lot of fun pushing it on our blog and asking the experts to really classify it. (Dave and John laugh) You know, what are we looking at? Well, you see a fin, is it a shark? Is it a dolphin? Is it a whale? It can get quite heated. >> I hope it's a dolphin, I hope it's a dolphin. (Sebastien laughs) All right, so, I want to get into the technology, but I'm just thinking about the practical operation of this. They're wind-powered. >> Sebastien: Yes. >> But they just can't go on forever, right? I mean, they have to touch down at some point somehow, right? They're going to hit water. How do you keep this operational when you've got weather situations, you've got some days maybe where wind doesn't exist or there's not enough there to keep it upright, keep it operational, I mean. >> It's a very good question. I mean, the ocean is often described as one of the toughest environments in the universe, because you have corrosive force, you have pounding waves, you have things you can hit, marine mammals, whales who can breach on you, so it's a very hard problem. They leave the dock on their own, and they sail around the world for up to a year, and then they come back to the same dock on their own. And they harvest all of their energy from the environment. So, wind for propulsion, and there's always wind on the ocean. As soon as you have a bit of pressure differential, you have wind. And then, sunlight and hydrogeneration for electrical power, which powers the onboard computers, the sensors, and the satellite link that tells it to get back to shore. >> It's all solar-powered. >> Exactly, so, no fuel, no engine, no carbon emission, so, a very environmentally friendly solution. >> So, what is actually on them, well, first of all, you couldn't really do this without the cloud, right? >> That's right. >> And maybe you could describe why that is. And I'm also interested in, I mean, it's the classic edge use case. >> Sure, the ultimate edge. >> I mean, if you haven't seen Sebastien's keynote, you got to. There's just so many keynotes here, but it should be on your top 10 list, so Google Saildrone keynote AWS re:Invent 2019 and watch it. It was really outstanding. >> Sebastien: Thank you. >> But help us understand, what's going on in the cloud and what's going on on the drone? >> So it is really an AWS-powered solution, because the drones themselves have a low level of autonomy. All they know how to do is to go from Point A to Point B and take wave, current, and wind into consideration. All the intelligence happens shoreside. So, shoreside, we crunch huge amounts of datasets, numerical models that describe pressure field and wind and wave and current and sea ice and all kinds of different parameters, we crunch this, we optimize the route, and we send those instructions via satellite to the vehicle, who then follow the mission plan. And then, the vehicle collects data, one data point every second, from about 25 different sensors, and sends this data back via satellite to the cloud, where it's crunched into products that include weather forecasts. So you and I can download the Saildrone Forecast app and look at a very beautiful picture of the entire Earth, and look at, where is it going to rain? Where is it going to wind? Should I have my barbecue outside? Or, is a hurricane coming down towards my region? So, this entire chain, from the drone to the transmission to the compute to the packaging to the delivery in near real time into your hand, is all done using AWS cloud. >> Yeah, so, I mean, a lot of people use autonomous vehicles as the example and say, "Oh, yeah, that could never be done in the cloud," but I think we forget sometimes, there are thousands of use cases where you don't need, necessarily, that real-time adjustment like you do in an autonomous vehicle. So, your developers are essentially interacting with the cloud and enabling this, right? >> Absolutely, so we are, as I said, really, the foundation for our data infrastructure is AWS, and not just for the data storage, we're talking about petabytes and petabytes of data if you think about mapping 70% of the world, right, but also on the compute side. So, running weather models, for example, requires supercomputers, and this is how it's traditionally done, so our team has taken those supercomputing jobs and brought them into AWS using all the new instances like C3 and C5 and P3, and all this high-performance computing, you can now move from old legacy supercomputers into the cloud, and so, that really is an amazing new capability that did not exist even five years ago. >> Sebastien, did you ever foresee the day where you might actually have some compute locally, or even some persistent-- >> So on the small Saildrones, which is the majority of our fleet, which is going to number a thousand Saildrones at scale, there is very little compute, because the amount of electrical power available is quite low. >> Is not available, yeah. >> However, on the larger Saildrone, which we announced here, which is called the Surveyor-- >> How big, 72 feet, yeah. >> Which is a 72-foot machine, so this has a significant amount of compute, and it has onboard machine learning and onboard AI that processes all the sonar data to send the finished product back to shore. Because, you know, no matter how fast satellite connectivity's evolving, it's always a small pipe, so you cannot send all the raw data for processing on shore. >> I just want to make a comment. So people often ask Andy Jassy, "You say you're misunderstood. "What are you most misunderstood about?" I think this is one of the most misunderstood things about AWS. The edge is going to be won by developers, and Amazon is basically taking its platform and allowing it to go to the edge, and it's going to be a programmable edge, and that's why I really love the strategy. But please, yeah. >> Yeah, no, we talked about this project, you know, Seabed 2030, but you talked about weather forecasts, and whatever. Your client base already, NASA, NOAA, research universities, you've got an international portfolio. So, you've got a whole (laughs) business operation going. I don't want to give people at home the idea that this is the only thing you have going on. You have ongoing data collection and distribution going on, so you're meeting needs currently, right? >> That's right, we supply governments around the world, from the U.S. government, of course, to Canada, Mexico, Japan, Australia, the European Union, well, you name it. If you've got a coastline, you've got a data problem. And no government has ever come and told us, "We have enough ships or enough data on the oceans." And so, we are really servicing a global user base by using this infrastructure that can provide you a thousand times more data and a whole lot of new insights that can be derived from that data. >> And what's your governance structure? Are you a commercial enterprise, or are you going-- >> We are a commercial enterprise, yes, we're based in San Francisco. We're backed by long-term impact venture capital. We've been revenue-generating since day one, and we just offer a tremendous amount of value for a much cheaper cost. >> You used the word impact. There's a lot of impact funds that are sort of emerging now. At the macro, talk about the global impact that you guys hope to have, and the outcome that you'd like to see. >> Yeah, you know, our planetary data is all about understanding things that impact humanity, right? Right now, here at home, you might have a decent weather forecast, but if you go to another continent, would that still be the case? Is there an excuse for us to not address this disparity of information and data? And so, by running global weather model and getting global datasets, you can really deliver an impact at very low marginal cost for the entire global population with the same level of quality that we enjoy here at home. That's really an amazing kind of impact, because, you know, rich and developed nations can afford very sophisticated infrastructure to count your fish and establish fishing quarters, but other countries cannot. Now, they can, and this is part of delivering the impact, it's leveraging this amazing infrastructure and putting it in the hands, with a simple product, of someone whether they live on the islands of Tuvalu or in Chicago. >> You know, it's part of our mission to share stories like this, that's how we have impact, so thank you so much for-- >> I mean, we-- >> The work that you're doing and coming on theCUBE. >> This is cool. We talk about data lakes, this is data oceans. (Dave laughs) This is big-time stuff, like, serious storage. All right, Sebastien, thank you. Again, great story, and we wish you all the best and look forward to following this for the next 10 years or so. Seabed 2030, check it out. Back with more here from AWS re:Invent 2019. You're watching us live, right here on theCUBE. (upbeat pop music)