2026
Live demo
Runbook Relay
A deterministic incident-response control room where a human and an AI agent share evidence, approval state, and an audit log.
WebMCPhuman approvalagent controls
View project Enterprise AI · Architecture · Reliability
I help enterprises move agentic AI from prototype to reliable production.
I am a Senior Solutions Architect in Munich. I work on agent reliability, evaluation, identity, policy controls, observability, and human approval.
ArchitectureAgent reliabilityEvaluationGovernance
Selected work
Systems you can inspect
2026
Reference lab
Mistral Reliability Lab
An inspectable Python reference for bounded tool use, provider switching, resilience, observability, and deterministic evaluation.
MistralOpenTelemetryPython
View project Technical writing
Ideas built for practice
Context and Control Are Different Layers
An agent needs enough context to reason and enough control to act safely. Treating them as one system hides different failure modes.
Read articleFrom Screenshots to Governed Tools
Agent-friendly interfaces should expose structured operations without taking the human operator out of the loop.
Read articleAn Ontology Is More Than Context
Context is the evidence available now. An ontology defines what the entities, relationships, and rules in that evidence mean.
Read articleTalks & workshops
Make the architecture discussable.
I develop practical talks around production-agent architecture, reliability, evaluation, and governance. Reviewed talk pages and materials will be added here as they become available.
See available topics