I built an AI ingredient scanner. no barcode, no database
Most ingredient scanner apps like Yuka work great — but only for Western products on a global barcode database. If you're in India and scanning a local brand, you get nothing.
So I built IngrediScan. Instead of a database, it uses AI vision to read the ingredient list directly from a photo and analyze it. You get a safety score out of 10, each harmful ingredient flagged with a plain-English reason. Works on food, skincare, supplements — anything with a label. Including brands nobody outside India has ever heard of.
Stack: React + Vite + Tailwind, Supabase for caching and waitlist, Claude Haiku for the AI analysis, deployed on Vercel.
3 free scans, no account needed: ingrediscan.in
Happy to answer questions about the build.
I'd be careful treating this as a database problem too quickly.
The interesting question may not be whether AI can analyze ingredient labels more effectively.
It may be what decision users are actually trying to make when they scan a product in the first place.
Those sound similar, but they can lead to very different conclusions about what gets prioritized, what validation matters, and where the real value sits.
I wouldn't make that call casually from early usage data.