Selected work
AI product work, built and shipped.
Everything here I built myself, from the product thinking to the prompts, evals, and deployment, first with no-code AI (Airtable, Softr, Zapier), now directly in Claude Code. Some are live, some are prototypes, some are early concepts. Open any card for the problem, the bet I made, and what I learned.
My process
From fuzzy problem to confident launch
Two frameworks anchor how I get there, whatever I'm building.
- Map the problem space
Opportunity Solution Tree
Connect the outcome to real customer problems, then to competing solutions, so the team commits to the right problem before anyone builds.
- Stress-test the plan
Pre-mortem
Assume the launch failed, then trace back why. Surfaces risks and hidden assumptions while they're still cheap to fix.
Then I build, measure, and learn, which is what the experiments below are.
LivePrototypeConcept
More AI Experiments