Spheer helps provinces monitor nature from space using AI

"I worked at Philips where we built machines that measured down to almost the micrometer," says Mark Boer, co-founder of Spheer. "Now I work with satellite images where one pixel equals 10 meters." From the smallest measurable detail to the big picture; Mark deliberately made the switch. Together with three co-founders, Jakko de Jong, Dominique Cirkel and Mark de Wijk, he founded Spheer, a company that trains AI models on satellite data. Their platform helps provinces monitor nature without needing to be AI experts themselves.

‍

Pretty quickly we thought: investment, that's not going to happen. We're going to do what we're good at first, namely projects and consultancy.

‍

Four founders, one leap

Mark knew his co-founders from an IT company in Groningen. "We worked as a small team on data science, where we were experimenting with satellite images and AI a lot." In the summer of 2021, they decided to take the leap together. "We decided to go full-time right away. The direction was clear because we wanted to combine AI and satellite images. How and what exactly we were going to do with it? That part was still open."

The four founders each came together from their expertise. Mark Boer brought his years of experience in micrometer precision from Philips, where he worked on measuring equipment for the chip industry. Jakko de Jong is responsible for Spheer's long-term vision and strategy.

Dominique Cirkel ensures that Spheer's technology aligns with users' needs and that projects both run smoothly and on schedule. Mark de Wijk, a self-taught developer who originally studied Russian before diving into IT, built the first application as a developer. The latter left the company in the summer of 2025.

Initially, they considered external financing, but soon chose a different route. "We had confidence in our abilities and in selling ourselves. Pretty quickly we thought: investment, that's not going to happen. We're going to do what we're good at first, namely projects and consultancy."

‍

Interpreting shades of green

"Just before we started, the EU launched the Sentinel-2 Copernicus program. Satellites whose data was freely available. Our timing was perfect. Each pixel of this satellite equals 10 meters in reality. Essentially, you're looking at many shades of green. We worked with textured images, but with many of those images per year. Because you look at the same area so often, you can track growing seasons well and see subtle changes in nature that the human eye misses. That's where AI is fortunately good at." 

‍

We had to build a custom AI model for each client. That doesn't scale.

‍

Seaweed and the turning point

The Province of Groningen became one of their first clients. "We monitor seaweed in the Groningen salt marshes. These are vegetated pieces of land that border the sea directly, without dunes. The bit of nature outside the dike." The province wanted to see annually what the impact of subsidies was. "This started as a pilot, but resulted in our first annual subscription."

But Mark saw a fundamental problem. "It takes quite a bit of expertise to train these models. You need a whole data science team." This meant their approach only worked for large clients with big budgets. "We had to build a custom AI model for each client. That doesn't scale."

The solution came when they won the SBIR competition from the Netherlands Space Office in 2023. "We got a research grant to develop geospatial foundation models." The idea: one basic model that ecologists can train themselves without technical knowledge. "The results were so positive that we decided to pivot our entire company. We created a platform where the client trains the AI themselves, rather than us training for them.

Thus Spheer was born. After eight months of alpha phase, the platform went public in 2025. "Now an ecologist without AI knowledge can train their model within a few hours."

‍

Spheer is much more accessible.

‍

Google comes around

The competition doesn't sleep. "Last summer Google launched a similar tool, a month later Cambridge too," says Mark. "It was a bit of a shock, but their launch confirms we're on the right track. The difference is that Google Earth Engine requires a data scientist. We focus on domain experts: ecologists who know a lot about their field, but not how AI works. Spheer is much more accessible."

‍

Measuring nitrogen without seeing it

Another practical example. A major Dutch problem is nitrogen deposition. "We model that because it's difficult to measure. With Spheer you can see the impact over the long term, allowing you to understand the effect on nature."

"Certain species are indicative. In heathlands, nitrogen leads to molinia grass, a grass species. By measuring these grasses, we see the impact of that nitrogen. If we see this grass increasing, we know there's an abundance of nitrogen present in the area."

‍

Change detection especially opens many new doors for us

‍

Change detection: the holy grail

"I'm convinced that in the future, much satellite image analysis will be done by AI." But there's more. "A major interesting development is change detection. Now we give annual updates on changes in satellite images, but many governments want to know when something changes. Think of a farmer illegally digging a ditch or building a barn. That can also be seen with satellite data over the long term."

"Change detection especially opens many new doors for us," Mark explains. We now primarily work for the parties in charge of nature, which are obviously far fewer parties than would otherwise be available. This allows us to reach a much broader market."

‍

The business model evolves

"Our first version was one-size-fits-all. Now we see that smaller clients require a different approach. We had to figure out how to capture that in a business model. That's why we're moving to an area-based model: per square kilometre of satellite data per year, you pay a certain amount. Logical, in our eyes. If a client is interested in a larger area, it costs more computing power and thus more storage."

Spheer's clients are mainly governments. "We want to change and expand that. A logical step for us are consultancies and ecological consultants. Some clients see the value but can't manage it. They just want the output. Consultancies can help us with implementing Spheer, so we can focus on developing the tool."

‍

In five years I hope for global embeddings and that people worldwide use Spheer. Change detection will also be fully launched by then.

‍

From Groningen to the world

This year Spheer is focusing on expansion. "Geographically, we train on data from all of Europe. That's why we want clients in Europe. We're also looking for a sales employee and an investment in the company," Mark outlines the situation.

"In five years I hope for global embeddings and that people worldwide use Spheer. Change detection will also be fully launched by then. Finally, I hope our data will be interactive by that time. You'll then get a direct answer to a question like: 'Where have people built solar panels within 10 kilometres of Groningen?'"

‍

Pick up the phone and talk to an entrepreneur. I was surprised how open everyone is.

‍

The constant uncertainty

"Sometimes I think: a salaried job is easier. You're constantly in uncertainty, but at the same time you learn a lot. Hiring the first employee was intense. For them, you suddenly become their safe place; you're responsible for everything. By the third employee you think: oh, another one. You get used to it quickly. And then something new comes along."

Want to break through the rut if you're stuck? "Pick up the phone and talk to an entrepreneur. I was surprised how open everyone is. They all have this mentality: we may not be working together now, but who knows what the future brings."

From seaweed in the salt marshes of Groningen to satellites over all of Europe, and from micrometers at Philips to 10-meter pixels. Mark and his team are building something big with Spheer, but they're doing it pixel by pixel.

‍

Interview & text by Lars Meijer