Current focus — updated as priorities shift.
Building
- Production data and AI systems through client work, including aviation computer vision and real-time insight pipelines
- Public tools for high-throughput observability and package inspection: LogWatchr and AptPeek
- An AI gateway for routing across hosted and self-managed models, with attention to caching and observability
Operating
- A Nomad and Consul cluster on dedicated hardware in the QTS Datacenter in Groningen
- Live public projects such as LogWatchr and AptPeek under ordinary production constraints
- Personal infrastructure kept deliberately simple: repeatable delivery, clear ownership, minimal ceremony
Exploring
- NVIDIA DGX Spark workflows for reproducible inference and embedding experiments
- Model-routing patterns that stay understandable when traffic, cost and latency pull in different directions
- Go and ClickHouse designs for high-throughput paths before they land in larger platforms
Improving
- Performance and reliability of systems already in production
- Observability that answers operational questions without turning into noise
- Feedback loops that shorten the distance between a change and evidence it worked
Writing
- Notes on implementation decisions, architectural trade-offs and incidents from operating data, AI and infrastructure systems
- Longer pieces when a topic needs more room than a note allows
Last updated: 29 July 2026. This page follows the nownownow.com idea of a public “what I’m doing now” update.