NutriGo
Decode Your Food, Redefine Your Health
The problem NutriGo solves
NutriGo demystifies packaged food nutrition for everyday Indians.
In a country where 101 million people live with diabetes and childhood obesity is surging, understanding what we eat has never been more critical. Yet food labels remain cryptic—hidden sugars disguised under 50+ names, confusing ingredient lists, and misleading health claims leave consumers in the dark.
NutriGo solves this by:
- Instant AI-powered scanning of any packaged product to decode sugar levels, calories, and hidden additives in seconds
- Smart Health Scores that cut through marketing jargon with evidence-based ratings (0-100 scale)
- Personalized alternatives suggesting healthier swaps based on your dietary goals whether managing PCOS, diabetes, or fitness targets
- Progress tracking with analytics that show how your food choices impact long-term health
No more squinting at fine print or Googling ingredient names in store aisles. Scan, understand, decide—all in under 10 seconds. Perfect for busy parents checking school snacks, fitness enthusiasts tracking macros, or anyone reclaiming control over what enters their body.
Challenges we ran into
Building NutriGo taught us that AI accuracy and real-time performance don't come easy.
- OCR Accuracy on Indian Packaging
Early prototypes struggled with multilingual labels (Hindi/English mix), curved surfaces, and poor lighting. Our Gemini AI integration initially misread "Sugar: 25g" as "Sugar: 2.5g"—a 10x error that could mislead diabetics.
Solution: We fine-tuned our image preprocessing pipeline with contrast enhancement and implemented multi-pass verification where critical values (sugar, calories) are cross-checked against nutritional databases. Added manual correction flags for edge cases.
- Real-Time Analysis Without Lag
Processing high-res images through Gemini AI while maintaining <3-second response times proved brutal. Initial builds took 15-20 seconds, killing the "instant insight" promise.
Solution: Implemented intelligent image compression (reduce resolution while preserving label clarity), server-side caching for frequently scanned products, and deployed on Vercel's edge network for lower latency across India.
- Health Score Algorithm Calibration
Our first scoring model was too harsh rating 90% of products below 50/100, making everything seem "bad" and discouraging users.
Solution: Recalibrated using FSSAI guidelines and real user feedback. Now scores reflect realistic benchmarks: <5g sugar per 100g gets bonus points, while artificial additives trigger penalties. The result? Actionable scores that guide choices without inducing guilt.
Tracks Applied (1)
Best Innovation
Technologies used