MESSY SYSTEMS?
LET'S FIX THAT.
A portfolio of my work and projects.
Every system here starts with something messy — an inbox, a spreadsheet, a tender, a wall of due dates — and ends with a human who only reads the answer.
- 01A vault that files itselfKnowledge
- 02An agent that texts firstAmbient
- 03A tester that proves a dashboard, dailyPractice
- 04A tender desk that scores people seat by seatProduct
- 05A schedule that plans itselfProduct
Same idea, five times. The interesting part is never the model — it is the plumbing that decides when the model is allowed to cost something, and what happens to the answer afterwards.
The Vault
A schema-disciplined Obsidian vault an AI keeps tidy — explore it as a living knowledge graph. The brain Hermes feeds.
OpenHermes
An always-on agent in one WhatsApp thread — crons brief you, monitors ping you, five in-code routers file what you send back. A day in its life, animated.
OpenAgentic Testing
Every morning an agent works through a live product the way a careful user would — every number against its source, every screen for what misleads — and the findings go to the developers with a working mock-up of the fix. A case study, told as a set of case files.
OpenTenderIntel
One dashboard for every live bid — AI reads the tender, scores your people seat by seat, and tracks compliance to a go / no-go. Live in production.
OpenDeadlines
An academic command center for a full course load — Claude reads your schedule, heat-maps every due date, and breaks each brief into a three-day plan. Built from scratch as a single file.
Open