Currently
A snapshot of what I'm working on, reading, and thinking about. Updated when it stops being true.
Last updated: September 2026
Skilling up
In a deliberate self-learning stretch — observability for agentic systems, Kubernetes, and taking RAG-Lab to production depth. Still happy to hear about senior AI/ML engineering and technical leadership roles — based in Hamburg, open to hybrid or remote across Europe.
The main thing
Observability — making agentic systems measurable instead of merely plausible.
In practice that means OpenTelemetry as the emission boundary, with Prometheus, Loki and Tempo behind it. The tooling is the easy half. The interesting half is the design calls underneath: which dimensions can safely become metric labels and which will quietly detonate your cardinality six months later, and whether trace context actually survives an async agent boundary. Get that second one wrong and the traces are decorative — you have paid for the instrumentation and can still only guess why a run went sideways.
It is the least glamorous layer of an agent stack and the one that decides whether you can make a claim about your system or just an assertion. RAG-Lab is the proving ground.
Learning
OpenTelemetry + the Grafana stack
Running the whole stack locally to find out where each piece actually earns its place.
Kubernetes
Getting past docker-compose and into how things actually run once they are someone else's problem at 3am.
Reading
AI Engineering — Chip Huyen
Reading it properly, not skimming for the parts I already agree with.
Watching
F1 2026
The regulation reset made this the most interesting season in years.
Before this
- April 2026 Deep in agent-platform work, shader experiments, and expanding the n8n template library.