I build products end to end: the schema, the service, the interface, and increasingly the AI in between. At Sourcerie I lead product engineering for a customer insights platform, and build the internal AI tools that keep a small team fast. This site is an editor: select things, move them about, make it yours.
The whole productI'm happy at any layer of the stack, which mostly means I don't sit around waiting to be unblocked. A normal day might be a Postgres schema and some fiddly Kysely, a React and Next front end, and the small interface details nobody remembers to put in a ticket. Owning all of it is usually why the seams end up lining up.
AI in productionI've built five internal AI tools at Sourcerie, and the demos further down are tidied-up versions of them. The hard part was never calling a model, it's making it dependable: coping with the strange things real users type, staying inside a token budget, and pushing slow work onto queues so nothing hangs.
Roadmap and the codeMost weeks I'm split between the roadmap and the code. I set the quarterly direction and feed into product strategy at the higher level, triage the user stories as they come in, and build the internal workflows the team runs on, then keep sharpening them so we ship more with less faff. I like staying close enough to the implementation that the strategy has to stay honest.
Obviously, I've built this whole thing with Claude Code (although the things I've done here are all real). The other user here, Jack, is an illusion, don't trust anything he says.
Current work
SourcerieThe customer insights OS for consumer brands on Shopify. Product engineering end to end: React, Node, Postgres, GCP, and everything between the schema and the shadows.
Tools I've built there, rebuilt here as simulated demos
Data ▶ demoAsk-the-database analytics: plain-English questions compiled into careful SQL.
Notes ▶ demoTurns transcripts, CSVs and recordings into designed webpages: a different layout, colour scheme and scorecard per note type, generated on the edge.
Pulse ▶ demoTracks how AI assistants talk about a brand: weekly sweeps across answer engines, scored against competitors, with cited quick wins.
Read Only Everything ▶ demoOne search across Slack, Gmail, Calendar, Notion and the database. Read-only scopes by design, so it can be trusted everywhere.
Slack botsreport-bug-bot and user-story-bot: Claude-powered bots that turn Slack threads into well-formed tickets and user stories in Notion, chasing missing context in-thread themselves.
Senior and lead product engineering roles, fractional or interim product leadership, and the occasional freelance build. The fastest routes are a comment on this file (the ＋ pin) or LinkedIn.