Workshops, proof-of-concepts and design partnerships for teams building scrapers and local LLM pipelines.

I build the AI and data systems I write about here.

Built in public

• A self-hosted newsletter stack on Hetzner: Listmonk, Docker, Caddy, double opt-in, deployed from CI. read the post
• A systematic Polymarket trading programme, posted week by week, green or red. read the series

Every system on this site is one I run myself.

How it works

• Workshop, half a day. I show your team how I work: the scraper that aggregates the information, and the local LLM that processes it. Live, in your office.
• Proof-of-concept, two to four weeks. One question, answered with running code and a number.
• Design partnership, ongoing. One piece of your pipeline built properly: information aggregation, or local LLM processing that keeps your data on your own hardware.