Senior engineer building AI-powered production systems. 20+ years across software and data engineering. Now focused on AI engineering: RAG, agentic workflows, and the data infrastructure underneath them. Based in Munich 🇩🇪.
- Retrieval systems: production RAG with evaluation harnesses, citation enforcement, GDPR-aware data handling
- Agentic AI: multi-model orchestration, tool use, human-in-the-loop checkpoints
- ML and data platforms: pipelines, feature stores, serving, drift monitoring
- Developer tooling: Unix-philosophy composable CLI tools in Python and Rust
Each project ships with a live demo, an architecture diagram, and real metrics once it's ready. Links go live only when they work: the first one already does.
| # | Project | What it is | Status |
|---|---|---|---|
| 1 | RAG over the EU AI Act | Citation-enforced RAG over the EU AI Act: hybrid dense + BM25/RRF retrieval, an offline eval harness with bootstrap CIs, FastAPI + Gradio demo |  |
| 2 | MarketFit Data Platform | Local-first data platform: bronze/silver/gold pipeline, structured skill extraction, evidence-backed profile scoring, RAG with citations |  |
| 3 | Job Research Brief Agent | Human-in-the-loop LangGraph agent that researches a company and drafts a tailored application brief, pausing for approval before anything is finalized | |
| 4 | Volatility Forecast Platform | Medallion-architecture platform: point-in-time feature store, volatility forecaster, FastAPI serving, drift monitoring | |
| 5 | Lossless Spec-Decode Bench | LLM inference and cost engineering: KV cache, continuous batching, speculative decoding, and quantization, verified lossless via a 300k-trial Monte-Carlo | |
| 6 | LLM Eval Leaderboard | LLM evals and observability: BLEU/ROUGE-L/F1/pass@k implemented from scratch with a bootstrap-CI leaderboard | |
Languages
AI / ML
Data
Infra & tooling
📍 Munich, Germany · 💼 Open to AI / Data / SWE roles (Munich / remote-EU)
