Self-hosted & reliable
Runs on infrastructure I administer myself — process supervision, health checks and auto-recovery so it stays up without babysitting.
AI infrastructure
I don't just use AI — I engineered the system that runs it.
Beyond shipping apps, I engineered and run my own AI infrastructure — a reliable assistant hosted on my own server, not a third-party wrapper. The interesting engineering isn't the model; it's everything around it: keeping it dependable, giving it a large persistent context, and structuring that context so it's actually useful to a human day after day.
Persistent context
A structured, versioned memory layer means it doesn't start from zero every session — it carries the state that matters forward.
Structured for human use
Context is organised and optimised deliberately, so the assistant is genuinely helpful for real day-to-day work rather than a novelty.
Directed, audited, owned
I design the architecture and write the specs the system works from; work is scoped, reviewed, and committed under human sign-off.
A precision instrument — directed, audited, owned.
This is infrastructure I designed and maintain, not a product built for me. It's the same discipline as any system I ship: clear architecture, versioned specs, and human sign-off on what matters.
See the projects it supports