This Leuven start-up builds AI in space

You can find the Dutch article in Bloovi here.

‍

Edgise is a young start-up from Leuven that puts artificial intelligence on top of satellites. The space industry is booming, which means that more and more satellites are being launched, which in turn generate more and more data. Sending that data to Earth is a major bottleneck. If Abraham won't come to the mountain, then the mountain should come to Abraham, thought the founders of Edgise who process data in space. Co-founder Nick Destrycker explains exactly how they do that and how the company, which saw the light of day in 2019, aims to grow from start-up to scale-up. "Within three years, we want to have our first solution in orbit," he says.

Under the wings of De Cronos Groep

"We make software and hardware. In doing so, we focus on edge computing or edge AI. In other words, we place artificial intelligence at the edge of the network. This is how we make the next generation of AI consume less and work more real-time," explains Nick Destrycker.

Instead of processing data in data centers, they do it where it is generated, being at the edge of the network. "We started doing that at the end of 2019," he echoes. "We are working under the wings of the Raccoons innovation cluster within De Cronos Groep. They are our incubator, so to speak."

"Initially, we focused on technology, and not on a market. Because we didn't have a market segment, we were all over the place," laughs Destrycker. "We were doing projects around smart cities, healthcare and food at the time. But we always had a limited impact. In the mid-2020s, however, we came into contact with space companies. That's how the ball got rolling, and we saw that our technology could have a big impact there. We entered into a collaboration with ESA, the European Space Agency. That went so well that we then made the confident decision to go full steam ahead with space exploration."

So, Edgise went from being a general edge computing start-up, to one with a strong focus on space. Was that pivot difficult? "It was very gradual," Destrycker answers. "We got in touch with a company that needed edge AI on a satellite. That's how we developed a demo, and the ball got rolling. We visited ESA together, and our first full-fledged space project saw the light of day."

"That focus now gives us a clear goal. All decisions made are entirely in the service of space exploration. It's easier to channel all your energy in one direction. That obviously creates certain risks, because it puts blinders on. But at the same time, that laser focus is incredibly important."

Detecting failures

Why is edge AI so useful in space, we'd like to know? Nick Destrycker: "The number of satellites being launched is growing exponentially. At the same time, the sensors on those satellites are becoming increasingly complex and accurate. That means more and more data is being generated in space. Which is good, because that drives a lot of innovation. But that data has to be able to be processed somewhere, and that while the processing capacity is not growing along with it."

"Today, all data in space are crudely transmitted to Earth to be processed in data centers. That's a huge bottleneck. We are at the limit of what we can transmit. Some data even goes unused as a result. Our solution is to implement algorithms, which are now in the data centers, on the satellites themselves."

‍

"Today, all space-based data are crudely transmitted to Earth to be processed in data centers. That's a giant bottleneck."

‍
The signals from satellites can interfere with each other, it reads. "If you have two people in a room, you can have a conversation very easily. But if there are 50 people in that same room, it will be much more difficult," Destrycker explained. "The same thing happens with satellites. We therefore create an algorithm so that the satellite itself can detect and identify disturbances, without having to do such a thing manually from Earth."

Space travel is a marathon

Edgise is currently still in its early stages. The young Leuven-based company just found its market, and is now trying to build sufficient experience in it. "We are in the early stages of our venture," Destrycker confirms. "We made a decision to focus on aerospace, a sector in which you can only grow by setting up smart collaborations. Collaboration in space in terms of edge AI, where co-optimization between hardware and software can make the difference, is crucial. With the intention of scaling up to a scale-up phase within a few years."

In other words, Edgise is poised to tighten their product-market fit in the coming months. "Space is a marathon, not a sprint," Destrycker argues. "As an entrepreneur, you have to be patient anyway, but in space, that patience is tested even more. That makes our growth slower, in a technological world where everything can change so quickly. But at the same time, this is a market where our passion lies, and where we can therefore hopefully make a lot of impact. "

‍

"As an entrepreneur, you have to be patient anyway, but in the space economy, that patience is tested even further."

‍
"The space economy is completely blossoming. The number of satellites keeps increasing. So, in the future, we expect a tipping point that could accelerate our growth quite a bit. Space is also learning more and more from start-ups. Today, we work for three years on a project to launch our technology on a satellite, in the past you easily had to wait ten years for that."

For now, they are funding themselves with money from De Cronos Groep, in addition to project grants from sources such as VLAIO and ESA. Venture capital is not in the cards yet. "There is currently no need to attract external capital," Destrycker states. "In any case, it is not the phase we are in today. Which is not to say that that may not change in the future."

Energy-efficient

Edgise's team also encounters many technical and practical challenges along their growth path. "Space is also enormously challenging on a technical level," Nick Destrycker confirms. "You can't just use an algorithm for terrestrial applications in space. You can't plug a satellite in, for example," he laughs. "There is a limited budget of energy, and AI does dare to consume a lot of energy. For that reason, we are constantly challenged to make the algorithms as small, compact and energy efficient as possible."

Bringing additional expertise on board is also incredibly difficult, according to Destrycker. "A lot of companies are currently facing that problem. It's already not easy to find someone who knows how to develop hardware for AI, let alone that person who also knows space. Therefore, finding the right profiles is a real struggle for us."

Nevertheless, they are very optimistic about the future at Edgise. "Within three years, we want to have our first solution in orbit," Destrycker makes a strong case. "That ambition is now our focus. In the next few years, we will finally really go into space!"