Team 30

SavingsDetective

A personal agent that finds post-purchase savings for lower income households

Check your own project

Community votes recognize the build and do not affect official judging.

Review the project

Start with the source code, then open the demo or video if available.

View GitHub repository
Project description
Andy is 45, father of two kids, Dayton, Ohio. He saved for months for a $700 TV. Two weeks later his friend got the same TV for $500 less. Best Buy's own policy says Andy is owed that $200. Nobody told him. Nobody ever does. SavingsDetective is a personal agent that reads every merchant's written policy, watches your orders, and pro-actively texts you the moment you're owed money: "Found: Best Buy dropped your TV by $200. Approve?" One yes. It then chats with merchant support, declines the gift card, cites the exact clause, and texts you when the cash lands. Every claim comes with a receipt: the clause, its source, the evidence, the full transcript, and a timestamped trail of every step. Rules decide the money. The model only writes the words. Built for the households with the least time to chase fees. The demo runs on real Amazon orders from a family of three earning under $25,000 a year (Open E-commerce 1.0, Harvard Dataverse) and finds $232 across six claims, including one it refused to file because the window had closed. Honest about what's fake: merchants are simulated and crash-tested with hostile support reps, eight for eight. Policies live in Senso. Events in Amplitude. The kill switch is a Statsig flag. Texts arrive over real iMessage. Instinct is a concierge for people who can afford one. SavingsDetective is a collections agent for everyone else.
Tools used