AI systems for industries where you're not allowed to get things wrong.
I build and deploy AI workflows inside regulated, multilingual environments — banking regulation, professional services, compliance. Working notes and systems below.
- Base
- Munich, DE
- Langs
- DE · EN · NL
- Domain
- FS · Regulation · AI
I'm a consultant in financial services, working at the intersection of data, AI, and regulation — currently in Munich, after 13 years across the Netherlands, Australia, Singapore, and the UK.
My work sits in a specific gap: organisations in regulated industries know AI matters, but most implementations die between the demo and the audit. I build the systems that survive that gap — structured AI workflows with clear operating rules, review layers, and human accountability built in — and I run the adoption side: getting real teams to actually use them.
I work in English, German, and Dutch, often in the same week. This site is not a blog. It's a set of working notes: systems I've built, what they do, and what I learned building them.
Each entry is a real system in active use — genericized where needed. No frameworks, no manifestos. Architecture, rationale, lessons.
- System 00
This Website
The site you are reading — spec'd in Claude, built by Claude Code, deployed in a day. The only system here whose output you can verify yourself.
- Langs
- EN
- Error tol.
- RECOVERABLE — git revert exists
- Review
- HUMAN SIGNS EVERY DEPLOY
- Status
- IN PRODUCTION
- System 01
Proposal Operating System
A six-mode AI assistant that runs a competitive tender end-to-end — from win strategy to pre-submission quality control — while holding state and discipline across a multi-week process.
- Langs
- EN · DE
- Error tol.
- ZERO — client-facing, deadline-bound
- Review
- HUMAN OWNS EVERY WORD
- Status
- IN PRODUCTION
- System 02
Auditable Document Workflow
An end-to-end pipeline that turns photographed paper receipts into verified, tracked financial records — built for a nonprofit where every entry represents real money and silent errors are unacceptable.
- Langs
- EN
- Error tol.
- ZERO — real money, real audit trail
- Review
- VERIFY BEFORE RECORD
- Status
- IN PRODUCTION