I build applied and agentic AI systems, with a particular interest in making them intuitive, intelligent, auditable and safe to give real permissions to. Previously: EU AI Act and AI governance work at the European Commission's DG DIGIT, and Go/Kubernetes engineering at SAP. Currently finishing an MSc at TU Dresden. Below: selected work, and occasional notes.
Recommended articles
Building a search that admits when it found nothingA journal is a bad corpus to search: bilingual, untitled, no shared vocabulary. Three retrieval arms, and a cut that is allowed to return zero results.Capture first: Building a journal with no filing stepA journal that never makes you file a thought before you've finished having it. Fully private and self-hosted.
All writing →