By day I build the control systems that keep a coral-reef restoration facility running at Coral Vita — .NET and React on the web, an Avalonia touchscreen on Raspberry Pi, all over MQTT.
After hours I build my own products end-to-end — Cronradar and HumanProof: scope it, build it, ship it, support it. AI is a tool, not the point.
Two roles, both close to the work: an unpaid IT internship at a vertical-farming company, and now full-time engineering at a coral-reef restoration company. Same instinct in both — own the problem, ship the fix.
I build the systems that keep a coral-reef restoration facility running: a .NET / ASP.NET Core web hub and an Avalonia touchscreen UI on Raspberry Pi that control water-treatment and per-zone life-support — pumps, valves, dosing and sensors — in real time over MQTT, deployed per site with Docker. C#, PostgreSQL and async background work.
Tech support, data migrations and system upgrades for a vertical-farming operations team — my first real foot in the door of a production codebase.
No new toys when the boring ones still work. AI is a tool, not a religion. Picked for joy of debugging, not novelty.
One live in production, one in active development — both scoped, designed, built and shipped by me, end to end.
Drop-in cron monitoring for .NET, Node, Python, PHP and more — heartbeat, schedule diffing and alert routing in about five lines. It grew out of the uptime monitoring I first built for surveyr.io.
It proves a real human did the work and measures how good they are — keeping only the score, never the recording. The model labels the evidence but never assigns the number, and raw capture is deleted the moment it's measured. Building toward a job board that ranks people by what they can actually do.
Building something where I'd scope, build and ship end-to-end? A role or fellowship that needs a utility player? A bug in one of mine? Drop a line — I read every email and reply to most within a day.