Evals and observability
Eval harnesses, RAGAS, Langfuse, production traces, and quality baselines.
Measured. Not vibes.
Hi, I am Johannes.
I make AI features production-ready for funded DACH startups: evals, retrieval, and cost control. Hands-on in your codebase, with numbers instead of guesswork.
Production beyond the model.
Three focused capabilities I can own with your team.
Retrieval and agents
Messy German-language data, RAG, orchestration, tools, and MCP integrations.
Cost and reliability
Inference economics and the unglamorous production wrapper around everything.
Selected experience
STIHL·MediaMarktSaturn·Porsche
Latest notes.
What I built, what broke, the numbers, and the fix.
You shipped the feature. Now let us prove it works.
Best fit: a funded DACH company under 200 people where the problem owner can sign the contract.
Book a technical fit call ↗