Independent AI engineering, from Amsterdam. Mostly retrieval-augmented and LLM systems at the moment: the ones that work in a demo and then have to survive real documents, real users and someone else maintaining them.
Retrieval that returns the right thing. Evaluation so you can tell when it stops.
Models out of notebooks and into something that runs, is monitored, and can be handed over.
Scraping, extraction and preparation, built to keep working after you stop watching it.
A short paid look at a system that is not doing what it should, before anyone commits to a rebuild.
MSc Artificial Intelligence at the University of Amsterdam, with a thesis on few-shot segmentation of CT scans at Amsterdam UMC. Before that, a computer science background and several years of building data-driven services for companies rather than for papers.
The research side is real and I keep it up, but the reason people hire me is the other half: I ship. See what I build or the research and writing.
Per project, with a defined deliverable, rather than an open-ended block of hours. For anything unfamiliar or already broken, I would rather start with a short paid diagnosis than quote a fix blind. It is cheaper for you and more honest than a guess.
I work remotely, on my own equipment, and I am straightforward about what I do not know yet. Rates on request, and they depend on the shape of the work.