Case Study
Intake
AI-powered nutrition and supplement scanner app that lets users scan any food or supplement label with their camera and instantly receive a personalized ingredient breakdown. Uses Gemini 2.5 Flash Vision for single-call label analysis, scoring products on Safety, Dosing, and Transparency at ~$0.002 per scan.
01
The Problem
Supplement labels are dense, misleading, and hard to evaluate. Most consumers can't tell if a product is properly dosed, contains risky fillers, or hides behind proprietary blends. Existing apps either require manual entry or give vague, unhelpful feedback.
02
The Approach
I built a camera-first mobile app that uses Gemini 2.5 Flash Vision to analyze supplement and nutrition labels in a single API call. The AI extracts every ingredient, cross-references dosing against clinical research, and returns a structured breakdown with Safety, Dosing, and Transparency scores — all in under 15 seconds.
03
Technical Highlights
- Single-call Vision API architecture — one image in, full structured analysis out, averaging ~$0.002 per scan
- Supabase backend with PostgreSQL for user scan history, saved products, and personalized flagging
- Custom prompt engineering to ensure consistent, structured JSON responses from Gemini across diverse label formats
- React Native + Expo for cross-platform mobile with native camera integration
04
Results
Currently in active development
Sub-15-second analysis time per scan
Cost-optimized to ~$0.002 per API call
Designed for scale with Supabase edge functions