Software | Social Technology Lab

Every day, more becomes possible.

What seemed out of reach yesterday
is within reach today.

Where do our opportunities lie?

Can we bringour peoplealong?

Where do we begin?

Not everything that is technically possiblemoves your organisation forward.

STL bridges the gap

between what technology makes possible

and what isworth buildingfor your organisation.

We explore the challenge with you, set the technical direction together and deliver, from first prototype to a reliable working system.

Which problems does STL take on?

What can STL build?

How do we decide what to build together?

1 · People first: We start with the organisation, its people and the problem. Before a single line of code, we invest in understanding what should be built, and why.
2 · Define the problem together: What are we actually solving, what would success look like and what is the shortest route to real value? Most of the thinking, and most of the gain, happens here.
3 · Design and build: Iteratively and close to real use. No grand reveal after months of silence: we build with your team, not behind closed doors.
4 · We stay through rollout: Projects tend to fail exactly where technology meets real processes and real people. We do not leave before that part is done.
5 · Build for independence: Documentation, training and internal capability: everything needed to carry on without us. Every engagement has a clear end.

How does your organisation stay in control?

Your call: We can build alongside your team, train your people, or take everything off your hands, including keeping the application secure and reliable in production. And you can change that arrangement later.
An independent lab: Social Technology Lab is independent, with no investors. Our business model does not depend on lock-in. You work with us because you choose to, not because leaving would be hard.

What does that look like in practice?

The challenge: Every week, the legal team had to review hundreds of new court rulings, up to two days of work.
Our approach: We first studied how the legal experts assess cases, then built an application that collects new rulings daily and lets AI pre-sort them.
The result: The system does the groundwork; the legal expert decides. Two days of work now takes a few hours.
What they can do now: The team now tracks all relevant case law, while we keep the application running on European infrastructure.

The challenge: Can in-house AI models improve weather forecasts? Answering that requires tens of terabytes of weather data, processed consistently every time.
Our approach: We began with a proof of concept: a pipeline capable of training a first model at that scale. Once it worked, we built the surrounding system with their developers.
The result: Reproducible datasets and a working foundation for training and refining their own models.
What they can do now: Their team carries the work forward on their own, and our involvement tapered off gradually.

The challenge: What does a language model trained on Dutch politics actually do? To study that transparently, parliamentary debates, letters and motions first had to become usable training data.
Our approach: We prepared two million political statements for fine-tuning, then trained two models that debate each other.
The result: A public debate between two models that shows policymakers both what the technology can do and where its risks lie.
Open to everyone: The code is public, so anyone can reproduce the work and build on it.

Stronger. Not more dependent.

You do not need to know yet what should be built. Tell us about the challenge you are facing.