I run a homelab, and most of what I make starts there: something falls over one too many times, so I build a tool to deal with it. I lean on AI where it genuinely helps, and I try to be honest about what I'm still figuring out.
What I actually work with
The honest version: what I reach for, what I'm building with, and what I'm actively learning. No proficiency bars, nothing I've touched once and listed as a skill.
Hands-on
Built real things with these
Building with
Shipping projects, still leveling up
Learning now
Where I'm headed next
Pinned projects
See all projects →net-triage
An LLM agent that reads a network log or device config and returns a structured, evidence-grounded diagnosis, pointing at the exact lines behind each call.
lazarus
A self-healing monitor for a homelab. It watches your hosts, shrugs off one-off blips, and restarts a downed service over SSH on its own.
agent-budget
A published npm package that tracks token usage and cost for LLM agents, so you can set a budget on an agent and actually see what it's spending.
Friday
A personal AI assistant built as a set of small agents that hand work off to each other, with a self-healing layer that brings things back when they fall over. Runs on my own hardware.
My GitHub
The last year of commits. Most of it is the homelab tooling and the projects above.