IT manager by day (22 sites, ~3,900 people). The rest of the time I build the automation: an AI-native helpdesk on Zammad, an LLM orchestration platform, and a K3s fleet of production agents, all on hardware I run.
- ttb-label-check: OCR verification of alcohol labels against COLA data. FastAPI, Tesseract, OpenCV. Live demo.
- chainward: on-chain investigation engine for AI agents on Base. Every numeric claim is re-verified before it publishes.
- zammad-ai-starter: Zammad plus an AI ticket-triage worker in one
docker compose up. - GPT-OS: self-hosted LLM orchestration, ~44K lines of Python, 630+ tests, in production since 2025. Private repo; architecture writeup.
- Homelab: 4-node Proxmox / K3s HA, Longhorn, Prometheus, GitOps, scheduled autonomous agents with least-privilege MCP access.
Python · TypeScript · Kubernetes · FastAPI · PostgreSQL · Claude / Ollama · Microsoft 365 / Entra ID
Writing: I replaced 1,000 lines of Python with a 500-word prompt



