What it is
Snagamon is a Pokémon-card marketplace I own and run. Collectors can browse and buy cards, and the standout feature is an AI-vision scanner: point a camera at a card and the system identifies it. It's a full production stack — modern web framework, database, authentication, payments-ready architecture — running live, with real users' expectations attached.
Why it matters for client work
A lot of AI consulting is slideware. Snagamon is the opposite: a shipped consumer product where an AI feature has to work on messy real-world input — bad lighting, worn cards, off-angle photos — or users bounce. Building and operating it is where a lot of my practical judgment about AI features comes from: what models can reliably do, where they fail, and how to design the product so failures are graceful.
It's also a working example of niche-marketplace fundamentals: structured product data, clean URLs, and the kind of category architecture that both search engines and AI assistants can navigate. The same patterns that make a card findable by a collector make it retrievable by a machine.