I work on making AI deployable inside environments that can't move to the cloud: energy utilities, industrial operators, KRITIS-regulated infrastructure.
Current focus
- On-premises LLM/RAG architectures under OT constraints (segmentation, no egress, auditability)
- The regulatory triangle for industrial AI: EU AI Act × NIS2 × IEC 62443
- Guardrail patterns for autonomous systems — invariants, human-in-the-loop, kill switches
📍 Frankfurt · 💼 Senior Consultant OT&ICS 🔗 LinkedIn
Repos here are personal work built from public sources — no employer or client material.
I’m building a local, multi-agent crypto research and paper-trading system in Python. It combines deterministic backtesting with bounded LLM-assisted research, point-in-time data, pre-registered hypotheses, leakage checks, locked holdouts, realistic costs, risk gates, and provenance.
The project is deliberately sandboxed: no live capital is deployed. Its first price/funding alpha study ended with a negative verdict, which is part of the research process—weak hypotheses are retired rather than continually tuned.