Yopido Adaptive AI

Yopido Adaptive AI helps people in Bolivia discover food and city plans that fit their preferences, and recover when a plan changes. Yopido was an existing restaurant SaaS before the event. The submitted contribution is a separate hackathon prototype: Pulse, Rewards and Guardian, not the pre-existing production platform. Pulse combines user-authorized preferences, menu candidates and weather context to recommend food and places. Rewards validates a campaign and its available budget before applying a discount to a test order. Guardian handles controlled incidents: an ordered item becoming unavailable, or simulated rain affecting an outdoor plan. Muse proposes an alternative, but cannot silently change an order or plan. The user must authorize the change, and the server revalidates stock, price and state before applying it once, with an audit trail. The demo uses real Meta model calls, test orders and explicitly simulated incidents. Weather context comes from MET Norway; simulated rain is not presented as an actual weather observation. Payment remains pending: no bank payment or reservation is claimed. The prototype runs with Next.js and local PostgreSQL and includes call, token and estimated-cost evidence and local consumption limits without automatic retries. The goal is to reduce repetitive coordination and help recover disrupted orders while preserving user and restaurant control. These are intended benefits, not measured business results. Full integration into Yopido production and a real-world impact pilot are future work.