Ingrechecker
Analysing food ingredients and identifying ultra-processed foods
The challenge
Ingredient lists are technically accurate and practically unreadable. The problem was turning inconsistent label data — from barcodes, open datasets and photographs of packaging — into an explanation a person can act on in a few seconds.
The product
A web application where a product can be identified by barcode or by photographing its label. Data from Open Food Facts is combined with OCR extraction and AI-assisted interpretation to classify ingredients and flag markers of ultra-processing.
The approach
- 01
Started as an experiment to test whether label interpretation was reliable enough to build a product on.
- 02
Combined multiple data sources with fallbacks, since no single source covers every product.
- 03
Used OCR to handle products missing from datasets, then normalised the extracted text.
- 04
Applied AI to explain ingredients in plain language rather than to produce a score with no reasoning.