I build backends in Go and Java, run them on Kubernetes, and write about what breaks when AI meets production.
Signals arrive by webhook, run through user strategies and go out as orders to several brokers. A Go backend, a SvelteKit frontend, and an AI layer that has to be right before anything executes.
biidin.comA Go backend, a desktop app with a live meeting copilot, an MCP server wiring LLMs into a real workflow, and live transcription. Every decision from architecture to shipping, owned end to end.
nevergoblank.comShort engineering notes from building both products. A model that answered and was thrown away, monitoring that had no callers, a retry that could never run, and what was changed each time.
privorum.com/insightsThirty years of Java releases, and how Project Lilliput cut the HotSpot object header from 12 bytes to 8 without touching a line of application code.
Read moreWhat is the Model Context Protocol, where it came from, and why every serious AI integration now speaks it.
Read moreWhy some software has to leave the browser — and how a thin bridge lets you keep web-quality UIs while running real native code underneath.
Read moreI take on backend work as a contractor: a service that has to be built, a legacy system that has to be modernised without stopping, or an AI feature that has to be made reliable before it ships. Remote, based in Porto.
Get in touchA project, a question about something I wrote, or a role. I read everything.