Every satellite image, aerial photo and point cloud that reaches a Terramira client has travelled a long way before it becomes a clear, actionable insight. To understand what that journey actually looks like, we sat down with Laurens Hagendoorn, Technical Lead at Terramira, who leads the developers turning raw earth observation data into Earth Intelligence.
From loose projects to one technical foundation
Terramira brought several specialisms together under one roof, and a large part of Laurens’ job is making sure that combined expertise actually reinforces itself technically. He is functionally responsible for around fifteen developers, safeguarding both the quality of what they build and the bigger architectural choices behind it.
“It’s easy to jump from project to project,” he explains. “But we only do projects well because we’ve also built technology that spans across them. There is a longer thread running through all the individual projects, so we keep getting better at serving all these individual client needs, built on one bigger foundation. If you’re not deliberate about that, you end up reinventing the wheel every time.”
You can build a great wheel, but it costs a lot of time. And then you keep building individual parts instead of an engine.
Laurens Hagendoorn, Technical Lead at Terramira
That longer line is exactly what Laurens spends most of his time protecting: setting up shared processes, aligning coding standards across teams that started from different places, and shaping how the team uses AI to help generate code consistently. Some parts of the system, like the API that serves the data, are held to a very high bar, thoroughly tested end to end, because any issue is felt immediately by clients.
One platform, one source of truth
Behind Terramira’s client-facing platform sits something more fundamental: an API through which nearly all analyses actually run. “In principle, everything a client can do in our platform, they could also build in their own frontend, simply by calling our API,” Laurens says. That is deliberate. The platform itself remains valuable because not every client has the technical setup to work directly with an API. It gives people an accessible way to see what is possible and, more importantly, to translate flat data into something they can actually base decisions on.
That same infrastructure and codebase carries very different services, from tree monitoring to agricultural monitoring to solar panel monitoring. “It all runs on the same system, the same way, the same codebase,” Laurens notes. “Consistent quality, and the same building blocks to get different results out of the same data.”
Teaching machines to see the land, and proving they see it well
A significant part of Terramira’s work relies on deep learning models trained to recognise objects such as trees, buildings, water and paved surfaces in satellite imagery, aerial photos and point clouds. Right now, the team is developing a foundation model, still in an R&D phase but expected to be usable by the end of the year, that will make it far easier to respond when a new satellite or sensor comes online, updating multiple services at once instead of retraining each one separately.
Alongside that, Laurens’ team is building validation datasets, manually labelled ground truth for elements like tree trunks and crowns, so the entire processing chain can be measured rigorously: from the raw output of the deep learning model to the domain rules applied afterwards by geospatial specialists, who bring in local knowledge, for instance recognising when something that looks like a tree in a particular setting simply isn’t one. Splitting that validation by landscape type, urban, rural or forested, means Terramira can state precisely how reliable its data is in each context, and demonstrate, with real numbers, how quality improves release after release.
Trees are this year’s main focus, currently Terramira’s largest service. By the end of the year, the ambition is to detect new trees annually from aerial imagery, alongside adding crown volume and finer species detection. It’s a good example of how an idea, tested first through student projects, gradually matures into a production-ready model that a colleague takes across the finish line.
Solving problems at a scale most companies never have to
Ask Laurens what makes Terramira’s technology distinctive, and the answer is direct: processing enormous volumes of data, very fast, for a large number of different clients at once. That means building systems flexible enough to serve each client on the same shared foundation, without starting from scratch every time.
We’re not Google. We don’t have infinite compute or infinite budget, so it has to be smarter.
Laurens Hagendoorn, Technical Lead at Terramira
It’s a philosophy shaped by experience. Earlier in his career, Laurens worked on a project for the World Bank, mapping electricity infrastructure across an entire country by predicting where the next pylon in a network was likely to stand, and buying only the imagery needed to confirm it, tile by tile, rather than purchasing satellite data for an entire country. The goal was to help identify where new solar investments would have the greatest impact on the grid, in places where that infrastructure often exists only as a drawing in a filing cabinet, not as digital data. He also built the earliest version of Terramira’s internal API for retrieving time series data per parcel, the same foundation that today underpins all of its agricultural monitoring.
Building the team as much as the technology
These days, what Laurens enjoys most is different: removing friction so developers can do good work and have a good day doing it. He is closely involved in hiring, building the team he now leads. Having started in his current role only a few months ago, he is candid that he is still learning fast, and that the room to grow is very much the point: it’s what allows Terramira’s technology, and the people behind it, to keep getting better.
Terramira gives me a lot of room to grow.
Laurens Hagendoorn, Technical Lead at Terramira
