The short version
I started building computers before I was a teenager — scrounged parts, tutorial videos, a first boot that posted by pure luck. That feeling of I built this never wore off. It just kept finding new things to become: PCs became servers, servers became a homelab, and the homelab became infrastructure for AI systems.
These days I build AI agent platforms with cognitive memory, ship products like Fish Bite that get used every day, and run a homelab that does everything from auxiliary ML inference to serving the whole stack. I'm not a company. I like understanding how things work from the ground up.
The longer version
The thing nobody tells you about building your first machine is that it rewires how you think. Once you realize a computer is components that snap together according to learnable rules, you start seeing everything that way. A server is a PC that doesn't sleep. An agent pipeline is software that follows rules you can inspect, test, and fix when they're wrong. The complexity is real, but it isn't magic.
I'm not a professional developer or a trained engineer. I'm someone who learns by building, breaks things regularly, documents what failed, and treats every rollback as tuition. The homelab has been rebuilt more times than I can count. The agent platform is on its seventh full rewrite. The catch log outvotes any model. That last part is the whole method: build, measure against reality, believe reality, rebuild.
What I'm actually good at
- Agent architecture — multi-stage pipelines, verification gates, unattended autonomy, memory systems that survive contact with real usage
- Local-first infrastructure — self-hosting as a lifestyle, right-sized hardware, services that survive 3 AM
- Scoring & modeling under real data — calibration against ground truth instead of vibes
- Cost engineering — multi-provider agent fleets, idle-subscription arbitrage, billions of tokens on consumer money
$ uptime
Still building. Still learning. No plans to stop.