Inspiration
Idaho’s 2026 SNAP rule (HB109) is hard to interpret at the shelf. Goal: instant, trustworthy eligibility from a barcode
What it does
Scans a barcode and returns eligibility, reason, and confidence. OFF-first product lookup (USDA FDC fallback), shows a product image when available.
How we built it
OpenCV barcode scanner Expo + React Native (camera via expo-barcode-scanner) calling a FastAPI backend. Node script fetches/normalizes OFF/FDC → piped to Python policy engine. Clear JSON contracts;
Challenges we ran into
Sparse data: “added sugars” and explicit “
50% juice” often missing.
Accomplishments that we're proud of
End-to-end flow (scan → fetch → normalize → policy) with fast responses. Deterministic, explainable decisions (no hallucinations). Modular design: swap data sources or policies without UI changes.
What we learned
Rules + evidence beat fuzzy heuristics for MVP accuracy and trust. Data provenance (OFF vs FDC) matters for debugging and confidence. Small, well-scoped scripts (download images, check with Python) speed iteration.
What’s next for Snap Scan
Caching + persistence, multi-state policy modules, and better explanations. Generalize to other states
Log in or sign up for Devpost to join the conversation.