The Problem
Packaged food labels are dense and technical — shoppers rarely have time to decode every ingredient mid-aisle. For people following Swaminarayan, Upvas, vegan, or vegetarian rules, a wrong purchase is more than inconvenience; it erodes trust in the product. Generic barcode apps surface nutrition facts, not principled compliance, and black-box AI answers feel unsafe when the decision has to happen in seconds.
Our Approach
- 01
Mapped the in-store moment — one hand, mid-aisle, seconds to decide — and implemented Scan → extract → analyse → verdict → save as the core mobile flow with minimal steps between camera and answer.
- 02
Built live camera capture with framing guidance on iOS and Android, optimised for quick retakes and low friction before analysis runs.
- 03
Pipeline: label image → text extraction → ingredient parsing → rule evaluation against the user’s selected dietary profile (Swaminarayan, Upvas, vegan, vegetarian).
- 04
Structured explainable results — per-rule verdicts plus which ingredients triggered pass/fail — so confidence comes from reasoning, not a single opaque score.
- 05
Shipped scan history as a persisted memory layer: repeat purchases get faster and the app supports habit, not one-off lookups.
- 06
Tuned latency and error states for store connectivity — clear loading, retry, and fallback copy when analysis is slow or incomplete.
- 07
Cross-functional delivery with product design over four months: engineering owned mobile, API, vision/LLM integration, and history storage; design owned in-store clarity and emotional reassurance.
Outcome
SatvikScan shipped on iOS and Android as a focused compliance companion: users move from label uncertainty to a clear, explainable verdict in seconds, with history that makes repeat shopping decisions faster and more confident.
