Blackbox

Sports media-rights deals are high-stakes and time-sensitive: knowing when a right comes up, and who holds it now, is most of the game. Blackbox fetches and visualizes that picture across European sport — rights, holders, and timelines — as a conversational surface backed by a Gantt-style view you steer by talking to it. It turns a research scramble into something a sports department can plan around, ahead of time.

The capability I built lets a user target a country, market, and sport and run a deep research pass. It’s a multi-step, self-validating chain: a broad search first, then every source re-checked against the original criteria, then targeted searches to fill the low-confidence gaps — repeating until the chain is complete and the results vetted, and only then the final synthesis. Searches run on Tavily and Perplexity Sonar; the two synthesis passes run on Claude; everything stays strictly true to the brief the client entered. I’d built this shape before — it’s the same validate-as-you-go chain I made for Centra’s course generation, redirected at sports-rights markets.

The scraper wasn’t on the brief. I proposed it in a meeting about business opportunities, because I could see what stacking this data unlocks: collect enough on who holds what and how deals get done, and the patterns become a sports-rights data vertical that’s arguably as valuable as the core business — in a niche, multi-million-dollar industry where it isn’t packaged anywhere. It also opens a next generation of features, fusing the scraper and its validation engine with the Gantt system already in place. The client asked for one thing; I delivered it, and added a new line of value on top.

Recruited as product designer and full-stack builder — I sharpened the entire experience (invited into the repo, building the changes AI-assisted and shipping them live) and built the self-validating market-scraper. The base product was a colleague’s build for the client; the design refinement and the scraper are mine.