Every studio is founded on a bet. Ours fits in one sentence: convenience is rented; control is built.

The rented path

The default way to add machine learning is to rent it. Data and prompts leave your perimeter for a cloud or a model API. Pricing, limits and behaviour change on someone else's schedule. Compliance rests on terms you cannot audit. Switching means rebuilding. For a demo that is the right trade. For a model that encodes years of process knowledge, or decisions the public is entitled to scrutinise, it means renting your own advantage back on terms you did not set.

What open weights changed

The rented path was defensible while the capable models sat only behind proprietary APIs. Open-weight models ended that. You can hold the weights, inspect them, fine-tune on your own data, run on your own hardware, and keep running with no deprecation notice and no behaviour change you did not approve. For the bounded problems production systems are made of, inspection, prediction, classification, document understanding, that is strong enough to build on. Ownership compounds; convenience depreciates.

Jurisdiction is an engineering requirement

GDPR, the EU AI Act and NIS2 are system requirements, as real as a latency budget. Where data may flow decides the architecture, what must be documented decides the pipeline, what must be auditable decides the logging. A compliance story that depends on a foreign vendor's terms is promised, not provable. So we decide data residency at the whiteboard, run models where the data lives, and produce evidence as we ship.

Why Gothenburg

Sweden's west coast built automotive, manufacturing and deep-tech engineering: systems designed to work for decades, safety margins calculated rather than hoped for, documentation part of the craft. That is what production machine learning needs. It is also where the work is: the factory floors and the public organisations we serve are here, and we would rather be thirty minutes from the machines and the municipalities than close to anything else.

In practice

  • Models run where your data lives. Your data never leaves the perimeter.
  • Open-weight models by default. Nothing critical behind someone else's API.
  • Full source code and documentation handed to your team, with an exit path at every layer.
  • GDPR and EU AI Act compliance designed in from the first commit, with evidence produced as we ship.
  • Every engagement ends in a system your own team runs.

That is the bet. The way to test it is small: a thirty-minute call and an honest answer about whether machine learning is worth it for your problem, on your own terms.