I help European companies get GenAI systems into production. Reliably, evaluated, and EU AI Act-compliant.
Most GenAI pilots die on the way to production: retrieval nobody measured, prompts nobody evaluated, compliance nobody scoped. I build the systems that survive that trip, and the harnesses that prove they work. AI Engineer at NFON AG, M.Sc. AI candidate, Google Cloud Professional Data Engineer.
- Based
- Ingolstadt, DE
- Now
- NFON AG
- Languages
- EN · DE · UR
- Consulting
- Limited availability
Selected work
Open source and production systemsspanscore ↗
An evaluation harness that scores RAG pipelines and agents straight from their OpenTelemetry traces: retrieval metrics, trajectory evaluation, LLM-as-judge with a calibration module that checks the judge against human labels, and a paired-bootstrap CI gate that blocks only statistically significant regressions. The benchmark is fully reproducible: committed traces and a judge cache re-score without an API key.
Enterprise RAG Platform ↗
Production RAG on Terraform-provisioned Azure: dense vector search over Azure AI Search plus cross-encoder reranking, served through FastAPI microservices with GDPR-compliant data handling.
sales-research-agent ↗
Provider-agnostic research agent on a custom MCP server over a mock CRM, with a code-enforced guardrail that every claim traces back to a real tool result. Claude and OpenAI swap behind one protocol.
GithubRepoAssistant ↗
Answers natural-language questions about any public repository. Pydantic AI with ten focused tools over the GitHub REST API that the model chains itself, so you can ask it things like “which open issues look stale?”
Work with me
Limited availability alongside my role at NFONGenAI Production Readiness Audit
One week, fixed scope, fixed price. I review the system you have and tell you exactly what stands between it and production: architecture, retrieval quality, evaluation, and the GDPR and EU AI Act gaps that surface late and cost the most.
You leave with a report you can act on with or without me. Built for German Mittelstand teams and the agencies delivering for them.
Enquire about the audit- 01Architecture & retrieval-quality review
- 02Evaluation plan: what to measure, and how you'll know it improved
- 03GDPR & EU AI Act gap list by risk class
- 04Written report with a prioritised fix list
- 0590-minute workshop with your team
Experience & education
2019 to today
AI Engineer
NFON AG · Remote, Germany
Data & AI Engineer
SII Technologies · Ingolstadt
Software Integration Engineer
e.solutions GmbH · Ingolstadt
M.Sc., Artificial Intelligence
TH Ingolstadt · part-time
B.Eng., Engineering & Management
TH Ingolstadt · grade 1.5 · Data Science, AI decision systems
Writing & speaking
Notes on getting GenAI past the pilotI measured my RAG system before believing it
Reranking looked like an obvious win. I built a calibrated eval harness around it, with the judge checked against human labels and a paired bootstrap on the deltas, and that is how I found out how much of the win was real.