FoodBridge

Inspiration

A teenage girl in Hong Kong went viral for spending her own money buying meal boxes from restaurants to hand-deliver to homeless people on the street. It was pure kindness, but entirely manual, unscalable, and dependent on one person's wallet. We were wondering: what if we could give every city that version of her, but streamlined? FoodBridge is the solution built for NYC, designed to turn everyday surplus into dignified meals, and make it effortless for anyone who wants to do good to actually do it.

What It Does

FoodBridge connects three groups who never had a shared platform. Restaurants post end-of-day surplus food in 60 seconds and receive an auto-generated donation record for tax reduction purpose for every completed pickup. Volunteers see listings sorted by distance from their live location, claim the nearest pickup, get auto-assigned the closest shelter drop-off, and verify delivery via Claude Vision photo recognition. Hungry people at NYC shelters receive fresh and hot meals that the existing food bank system, built for bulk scheduled deliveries, never reaches. Restaurant wins. Volunteer wins. Hungry people win.

How We Built It

React + Vite frontend, Firebase Firestore for real-time listings and volunteer data, Mapbox GL for the live pickup map, and the Claude API powering three distinct features: route optimization for multi-stop volunteer pickups, Vision-based delivery photo verification, and donation record generation using USDA reference pricing tied to real shelter EINs.

Challenges We Ran Into

The hardest product problem was incentivizing restaurants to take on extra logistics. Most food waste apps fail here, because they ask restaurants to do more work for no return. Our answer was the auto-generated tax donation record: every pickup produces a document an accountant can actually file, tied to a real 501(c)(3) shelter EIN, with fair market value calculated from USDA reference rates. Turning food waste into a documented tax asset changed the conversation entirely. On the technical side, integrating Mapbox with real-time Firestore listings and keeping marker state in sync with live claim updates required careful ref management across the map and sidebar components.

Accomplishments That We're Proud Of

We mapped the full three-sided product flow: restaurant, volunteer, and shelters/food pantry, and actually built it end to end in limited time. The donation record pipeline, the Claude Vision handoff verification, the nearest-shelter assignment logic, and the real-time listings map all work together as a coherent system. Most food rescue tools solve one side of the problem. We solved all three simultaneously.

What We Learned

Clear product flow mapping before writing a single line of code saves hours of refactoring. We also learned how to navigate merging changes across parallel codebases without breaking existing Firestore field contracts.

What's Next for FoodBridge

Three immediate priorities: First, the donation record is currently generated from hardcoded USDA reference prices — we want it pulling from live Firestore data per completed pickup, with a real PDF pipeline via Cloud Functions so it's tamper-proof and audit-ready. Second, restaurant authorization is currently hardcoded — the next version needs proper authentication so donation records are tied to verified legal entities with real EINs. Third, all seed data needs to migrate to live data: real restaurant geocoding via Google Places, real volunteer geolocation matching, and real shelter capacity feeds so the platform responds to what's actually available tonight — not what we hardcoded for the demo.