Directing AI, not just using it.
Four self-directed projects about working with AI tools: generating and revising media, prototyping software, designing a multi-role workflow, and automating a reporting pipeline. In each one I set the requirements, reviewed every output, and made the calls. The write-ups show the revisions, not just the results.
Fix the frame. Then make it move.
Approving the still before touching the video so mistakes don't carry forward, splitting a reveal that kept breaking, and catching the details that stop a shot from reading. Before-and-after at every step.
Read the case study →
Coverage Breaker
Football coverage rules turned into a working browser prototype. I defined the product, wrote the simulation logic and requirements, tested it, and directed the AI-assisted build. Playable demo included.
Read the case study →
Keeping several AI roles on the same facts
A research-heavy project used as the test case for a human-led AI workflow: how to divide work across multiple AI roles, keep them working from one shared set of facts, and stay the person making the decisions.
Read the case study →
Personalized Report Automation
Could one intake score a lead, shape a useful report, and point the follow-up in the right direction? An Airtable and Zapier build shown with live evidence, clearly separated from the parts that were only ever recreated. It never launched, and the write-up says so.
Read the case study →