A software-as-a-service platform for an underserved market: volunteer-run scout troops managing rosters, events, and advancement on decades-old tools. Owned the full product cycle solo, and used AI tooling extensively for competitive analysis, user insights and needs, product strategy, design, and coding of the full stack. Currently in free pilot with troops in active use.
Rather than guessing, I used an LLM to analyze and derive insights from over 10,000 email communications among Scouts, their parents, and troop leaders from a live scout troop. What surfaced was not what a feature list might have predicted: the volume was dominated by the same logistics repeated week after week, event after event — who is coming, what to bring, which form is still missing. That reading is what put events, RSVPs, and documents ahead of everything else in the build order. A close second was a modern way to track progress towards merit badges and achieving advancement.
Every record in this product belongs to a minor, so protecting that data was an architectural constraint designed with Claude Code and Supabase, rather than a compliance step at the end. The data model was designed around it from the first table: what a parent can see, what a leader can see, and what nobody outside the troop can reach are decisions enforced at the data layer, not in the interface. Building it in from the start meant the answer to “who can see this?” was never retrofitted onto a schema that had already assumed otherwise.
Troops do not organize themselves the way an off-the-shelf roster tool assumes. Scouts belong to patrols, patrols shift, adults hold several roles at once, and advancement is a tree of requirements rather than a status field. The user experience follows the real shape of the organization, which is why a parent, a scoutmaster, and a merit badge counselor each open the same product and see a view of what is important to them.