Ranked, not listed. The builds that carry the story come first, then the rest of the shelf. Each one is real, built to remove some manual work from my own life.
Large industrial project · field case study
A body of automations that turned the commissioning paperwork into software.
On a large minerals-processing expansion, my day was repetitive, high-stakes data work across 1,300+ subsystems and 14,000 equipment tags. I built a toolkit of automations plus an AI assistant that runs them. The biggest single win took the energization documents from typed by hand to generated from a template.
Personal AI assistant · v2, in daily use
A voice-first assistant with specialist agents behind a router.
A Next.js progressive web app, a Node worker running a queue of specialist agents behind a router, and a Postgres schema underneath, running unattended. Training, money, career and notes each get their own agent, and the arithmetic they need is done in SQL and handed to them, so an agent chooses and never has to remember. Otto was v1. This is what I use now.
finance.qori.land
A production data pipeline with an AI layer on real money.
Plaid pulls real transactions from multiple banks, Claude analyzes spending, and Supabase stores everything behind row-level security. A real, secure, multi-user data product with AI on top, not a toy demo.
Industrial commissioning · sub-case
The commissioning paperwork, done by three desktop tools.
Industrial commissioning · sub-case
Manual report and readiness-package work, replaced with code.
Personal build · indoor trainer
An indoor trainer controller that holds an ERG target over Bluetooth.
The rest of the story, the field work and the research, is on the front page.
Back home