[ AI DELIVERY LEAD · CONTRACT ]
AI delivery that survives contact with governance.
I find the process quietly eating a team's time, rebuild it with AI, prove the saving, then move it onto local, governed models so it runs at near zero cost with the data staying put. Most recently: AI in production within six weeks of joining a distressed programme, and a regulated healthcare AI system live with frontline emergency call handlers.
ABOUT
My test for any AI system is not how much it can produce, but what it is not allowed to do, and who signs it off. That governance discipline sits alongside delivery in everything below.
I cannot write code, and that is rather the point of this site. Everything in the project index below was built by describing outcomes clearly to AI tools, then governing what came back: testing it, documenting it, deciding what it is not allowed to do. It is the same discipline as the delivery work above, just applied to spare hours instead of a client's programme.
Project Index
04 ENTRIESLocal AI Job Hunter
A free, local-first job search assistant: it finds and scores roles, tracks every application, and drafts tailored CVs only for roles chosen to pursue. The guardrails are the product: it will not spray generic applications, it will not claim a skill that can't be defended in an interview, and nothing goes out unread.
VIEW ON GITHUBMinecraft with Claude Code: a parent's guide
A one-page guide to running a safe family Minecraft server at home, with an AI builder that turns "build us a waterpark" into a real one in twenty minutes. Real logins only, a guest list controlled name by name, every join logged.
READ THE GUIDEWing foil performance analysis
Six months of Garmin GPS data (around 110,000 points across 22 sessions) turned into a measured answer: longest unbroken flight went from 78 seconds in January to 23 minutes in June. Flight detection, speed analysis and offshore wind data, matched and visualised.
Second brain pipelines
Local, zero-cost pipelines that turn meeting audio and podcast archives into searchable, linked knowledge bases: automatic transcription with speaker recognition, then ingestion into a structured vault an LLM can query and maintain.
VIEW THE DEMOWRITING
Six Weeks of Discovery
Everyone was using AI. Nobody could say what problem it was solving.
Writing on LinkedIn about AI adoption from a delivery perspective: governance before headcount, measurement before spend, and what non-developers can now build for themselves.
FOLLOW ALONG ON LINKEDIN