 Okay, welcome back everyone to theCUBE's live coverage here in Las Vegas for Amazon Remars. Two days of coverage. We're getting down to wrapping up day one. I'm John Furrier, host of theCUBE. Space is a big topic here. You got machine learning, you got automation robotics. It all spells Mars. The two great guests here to really get into the whole GEO scene. What's going on with the data? We've got Marcus Norgren, business developer in GEO data. Sogetty, part of Capgemini Group and Joakim Welkist, portfolio lead data and AI with Sogetty, part of Capgemini. Gentlemen, thanks for coming on theCUBE. Appreciate it. Thanks for having us. So coming all the way from Sweden to check out the scene here and get into the weeds and the show. A lot of great technology being happening. Space is the top line here, but software drives it. You got robotics. A lot of satellite. You got the aerospace industry colliding with hardcore, industrial, I'd say IoT. Robotics want to put it whatever you want. But space kind of highlights the IoT opportunity. There is no edge in space, right? So the edge, the intelligent edge, a lot going on in space and satellite's one of them. You guys are in the middle of that. What are you guys working on? What's the focus here for Capgemini or Sogetty, part of Capgemini? I would say we focus a lot of creating business value, real business value for our clients with the satellite available. Actually a freely available satellite images working five years now with this solutioning and mostly in vegetation management and forestry. That's our main focus. So what's the product value you guys are offering? We basically, for now, the most value we created is working with a forest client to find bark beetle infestations in spruce forests. It's a big problem in European Union and the northern region of Sweden where we live. Now with the climate change, it's getting warmer. The bark beetle they swarm more times during the summer which makes it spread exponentially. So we help with the satellite images to get with data science and AI to find these infestations in time when they are small, before it's spread. So satellite imagery combined with data. This is the intersection of the data piece, the geo data, right? Yeah, you can say that. You have a lot of open satellite data and you want to analyze that but you also need to know what you're looking for and you need data to understand, in our case, a certain type of damage. So you have large data sets that we have to sort of clean and train ML models on to try to run that on that open data to detect these models. And when we're saying satellite data and open data, it's basically one pixel is 10 by 10 meters so it's not that you will see the trees but we're looking at the spectral information in the image and finding patterns. So it can actually detect attacks that are like four or five trees big using that type and we can do that throughout the season. So we can see how you start seeing one, two attacks and it's just growing. And then you have this big area of just damage. So how long does that take? Give me some scope to scale because it sounds easy. Oh, the satellite's looking down on us. It's not, it's a lot of data there. What's the complexity? What are the challenges that you guys are overcoming? Scope to scale. It's so much complexity in this. First you have clouds. So it's open data set. You download it and you figure out here we have a satellite scene which is cloudy. We need to have some analytics doing that, taking that image away basically or the section of the image which is cloudy. Then we have a cloud-free image. We can't see anything because it's blurry. It's too low resolution. So we need to stack them on top of each other and then we have a next problem to correlate them so they are pixel perfect overlapping so we couldn't compare them in time and then they have a histogram adjustment to make them like the sensitivity is the same on all the images because you have solar storms, you have shady clouds which could be used still that image. So we need to compare that. Then we have the ground truth data coming from a harvester for instance. We got 200,000 data points from the harvester. Real data points where they had found bark beetle trees and they pulled them down. The GPS is drifting 50 meters. So you have an uncertainty where the actually harvester was and then we had a crane on 20 meters. So you know the GPS is on the actual machine and the crane were somewhere. So you don't really know. You have this uncertainty. It's a data integration problem. Yeah. Massive. A lot of interesting things to adjust for and then you could combine this into one deep learning model and build a solution. On top of that, I don't know if you said that but you also get the data in the winter and you have the problem during the summer. So we actually have to move back in time to find the problem, label the data and then we can start identifying. So once you get all that heavy lifting done or write the code or I don't know what's going on there you get the layering, the pixel pack, see all how complex that is. When the deep learning takes over, what happens next? Is it scale? Is it all the heavy lifting up front? Is the work done up front? Is the scale on the back end? So first, the coding is heavy work, right? To get hands on and try different things, figure out in math how to work with this uncertainty and get everything solved. Then you put it into a deep learning model to train that. It actually run for 10 days before it was accurate. First iteration, it wasn't accurate enough. So we scrapped that, did some changes, then we run it again for 10 days. Then we have a model which we could use and interfere new images every day pretty quickly. Every day it comes a new image, we run it, we have a new outcome and we could deliver that to clients. Yeah, I can almost imagine, I mean the cloud computing comes in handy here. So take me through the benefits because it sounds like the old expression, the juice is not worth the squeeze. Here it is worth the squeeze if you can get it right. Because the alternative is what? More expensive gear, different windows, just more expensive monolithic solutions, right? Think about the data here, so it's satellite scene. Every satellite scene is 100 by 100 kilometers. So pretty much, right? And then you need a lot of these satellite scenes over multiple years to combine it. So if you should do this over the whole Northern Europe, over the whole globe, it's a lot of data. Just to store that, it's a problem. You cannot do it on prem. And then you should compute it with deep learning models. It's a hard problem if you don't have to. Well, you guys got a lot going on. So talk about Segeti, part of Capgemini. Explain that relationship because you're here at a show that you got, I can see the Capgemini angle. This is like a little division of the group. You guys like blown wolves. Like what's it like? Is this dedicated purpose built focus around aerospace? No, it's actually, so Segeti was the name of the Capgemini company from the beginning. And they relaunched the brand 2001 I think, roughly 20 years ago. So we actually celebrate some anniversary now. And it's a brand which is more local, close to clients, out in different cities. And we also tech companies. We are very close to the new technology, trying things out. And this is a perfect example of this. It was a crazy idea five years ago, 2017. And we started to bring in some clients, explore really open-minded, see can we do something on the satellite data. And then we took it step by step together with our clients. And it's a small team, we're like 12 people today. And you guys are doing business development. So you have to go out there and identify the kinds of problems that match the scope of the scale. So what we're doing is we interact with our clients, do some simple workshops or something and try to identify the really valuable problems like the Spruce Park people. That's one of those. And then we have to sort of look at, do we think we can do something? Is it realistic? And we will not be able to answer that 100% because then there's no innovation in this at all. But we say, well, we think we can do it. This will be a hard problem, but we do think we can do it. And then we basically just go for it. And this one we did in 11 to 12 weeks, a tightly focused team and just went at it. Super slim process and got the job done and the results were great. Well, it's interesting, you have a lot of use cases where you got to go down, do that face-to-face, belly-to-belly, body-to-body sales, business dev, scope it out of workshops. Now this market here, RIMARS, they're all basically saying call to arms. More money's coming in. The problems are putting on the table. The workshop could be a lunch meeting, right? I mean, because Artemis, there's a big set of problems to tackle. So I mean, I'm just oversimplifying, but that being said, there's a lot going on opportunity-wise here that's not as slow maybe as the biz dev that, you know, coming in. This is a huge demand. It will be explode. What's your take on the demand here, the problems that need to be solved and what you guys are going to bring to bear for the problem? Now we have been focusing mainly in vegetation management and forestry, but vegetation management can be applicable in utility as well. And we actually went there first, had some struggle because it's quite detailed information that's needed. So we backed out a bit into vegetation in forestry again, but still it's a lot of application in utility and vegetation management in utility. Then we have a whole sustainability angle. Think about the auditing of rogue harvesting or carbon offsetting in the future even by a diversity offsetting that could be used. And just to point out and give a little extra context, all the key knows talk about space as a global climate solution potentially, the discoveries and are also the imagery they're going to get. So we kind of got, you know, top-down bottoms up if you want to look at the world as bottom and space kind of coming together. This is going to open up new kinds of opportunities for you guys. What's the conversation like when this is going on? You're like, oh yeah, let's go in. What are you guys going to do? What's the plan going to hang around and ride that wave? I think it's all boils down to finding that use case that need to be sold because now we understand the satellite scene, they are there. We could, there is so many new satellites coming up already available that have come up. The cloud platform, AWS, it's great. We have all the capabilities needed. We have AI models needed, data science skills. Now it's finding the use cases together with clients and actually deliver on them one by one. It's interesting. I'd like to get your reaction to this, Marcus too, as well, your guys are kind of, you're a lot bigger and bigger than some of the startups out there, but the startup world, they find their niches and the workflows become the intellectual property. So your techniques of layering, how we'll see is an advantage out there. What's your guys view of that intellectual property of the future? Open source is going to run all the software. We know that. So software is going to open source. Scale and integration and then new kinds of ways are new methods. I won't say for just patents, but like just for intellectual property, differentiation. How do you guys see this as you look at this new frontier of intellectual property? That's, it's a difficult question. I think it's, there's a lot of potential if you look at open innovation and how you can build some IP which you can out license and some you utilize yourself and you can build like a layered business model on top so you can find different channels on markets we will not go for. Maybe some of our models actually could be used by others where we won't go. So we want to build some IP, but I think we also want to be able to release some of the things we do. Open works. Yeah, because it's also built presence. Community. Yeah, exactly. Because this problem is really hard because it's a global thing. And it's, imagine if you have a couple of million acres of forest and you just don't go out walking and trying to check what's going on. Because it's, you know. Manual's hard. Yeah, it's possible. So you need this to scale and it's a hard problem. So I think you need to build a community because this is, it's a living organism that we're trying to monitor if you talk about vegetation and forest. It's changing throughout the year. So if you look at spring and then you look at summer and you look at winter, it's completely different what you see. Yeah, that's interesting. And so I wonder if, you know, you see some of these crowd sourcing models on participation, you know, small little help. But that doesn't solve the big puzzle. But you have open source concepts. We had Anna on earlier. She's from the Amazon sustainability data project. And this is like open up the data. So it's data party for her. So in a way, there's more innovation coming potentially. If you can get that thing going, right? Get the projects going. And all this actually, our work, is started because of that. So European Space Agency, they decided to hand out this Copernicus program and the Sentinel Satellites, Sentinel-1 and 2, which we have been working with, they are freely available. It started back in 2016, I think. And because of that, that's why we have this work done during several years. Without that data freely available, it wouldn't have happened. I'm pretty sure. Well, what's next for you guys? Tell me what's happening. Here's the update. Put a plug in for the group. What are you working on now? What are you guys looking to accomplish? Take a minute to put a plug in for the opportunity. I would say scaling this. Scaling, moving outside Sweden, of course. We see our model that they work in US. We have tried them in Canada. We see that we work. We need to scale and do field validation in different regions. And then I would say, go to the sustainability area of this. There is a lot of great potential. International, too. It's huge. One area I think that is really interesting is the combination of understanding the carbon sink and the sequestration and trying to measure that. But also on top of that, trying to classify certain keystone species' habitats to understand if they have any space to live and how can we help that to sort of grow back again. Understanding the history of the forest. You have some data online, but trying to map out how much of this has been turned into agricultural fields, for example. How much of the real old forest that you have left that is really biodiverse? How much is just eight years young? To understand that picture, how can we sort of move back towards that blueprint we probably need to? And how can we digitize and change forestry and the business models around that? Because you can do it in a different way or you can do both some harvesting, but also not sort of ruin the whole ecosystem. They can be more efficient. You make them more productive, save some capital, reinvest it in better ways. And then you have robotics. And that's maybe something that we're not so active in, but I mean, starting to look at how can autonomy help forestry? Inventory damages, flying over using drones and satellites. You have people looking into autonomous harvesting of trees, which is kind of insane as well, because they're pretty big. But this is also happening. So I mean, what we're seeing here is basically... I mean, I've read a story multiple times called On Sale Drone, one of my favorite stories. The drones that are just like getting bobbed around in the ocean and they're getting great telemetry data because they're indestructible. You know, they can just bounce around and then they just transmit data. You guys are creating an opportunity, some will say problem, but by opening up data, you're actually exposing opportunities that never have been seen before because you're like, it's that scene where that movie Jody Frost of Contact where open up one little piece of information and now you're seeing a bunch of new information. You look at large scale data, that's going to open up new opportunities to solve problems that were never seen before. You can't automate what you can't see, right? That's the thing. So... We haven't even thought that these problems can be solved. It's basically, this is how the world works now. Because before, when you did remote sensing, you need to be out there. You need to fly with a helicopter or you put your boots on out and go out. Now you don't need that anymore. Which open up that you could be... You can move your creativity to another problem. Now you open up another problem space. So again, I like the problem solving vibe of the... It's not like, oh, catastrophic. Well, the earth is on a catastrophic trajectory. It's like, oh, we'll agree to that. But it's not done deal yet. Got plenty of time, right? So like, let's get these problems on the table. And I think this is the new method. Well, thanks so much for coming on. You really appreciate the conversation. Love it. Opening up new world opportunities. Challenges is always opportunities when you have challenges. You guys are in the middle of it. Thanks for coming on. I appreciate it. Thanks guys. Okay, Cap Gemini in the Cube. Part of Cap Gemini. So Getty part of Cap Gemini here in the Cube. I'm John Furrier, the host. We'll be right back with more after this short break.