Zuri turns a short description of a business into a live, on-brand website, matching it to the right template out of 80 and writing the copy to fill it in.
Zuri started as an entry for the Gemini X Prize. The brief was open: build something meaningful with the Gemini API. But the hard part wasn't the API itself, it was deciding what "building a website automatically" should actually mean. A generic AI website generator tends to produce generic, forgettable sites. I wanted Zuri to feel like it understood the business it was building for.
Rather than generating layout from scratch on every request, which is slow, unpredictable, and easy to get wrong, I designed a template-matching system instead:
This split the problem in two: a design problem (build 80 genuinely good templates) and a matching problem (get the AI to pick and fill correctly), both far more solvable than "generate any website from nothing."
Zuri is live at buildzuri.com with the builder running at app.buildzuri.com. It's currently at MVP stage, validated with test users but not yet an active user base, with ongoing work to broaden template coverage and refine how well the AI matches and writes for each business.