FAR let people with allergies or specific diets know whether a product was safe by pointing their phone at it. I co-founded it with two partners, one covering sales and strategy and one backend development. I owned product strategy and built the front app myself, both UX/UI and development.

The augmented reality was what people remembered from demos. It wasn't what made the product work. For someone with a real allergy a wrong answer is a health risk, so the actual challenge was coverage and accuracy: near-total product recognition, from launch. Great AR over a thin database would have been worse than useless. We solved it through data partnerships, using Open Food Facts for a large community-sourced base to launch with, then Alkemics for qualified manufacturer data direct from the likes of Unilever and Franprix. Around 100,000 products at launch, growing from there.

The design decision I'm most attached to came from taking that risk seriously: never show a bare yes or no. The app always displays the product it recognised, so on the rare false positive the user sees immediately that it scanned the wrong thing. For severe allergies I went further and warned users to double-check the packaging, since crowd-sourced data doesn't support a legal guarantee and pretending otherwise would have been the wrong call. Within those honest limits it stayed genuinely valuable, catching allergens and additives hidden behind codes such as E244, or buried in a language you don't read.

Thousands of downloads across iOS and Android in the first months confirmed the demand. The project won the Jury's Favourite Award at the 2016 FoodHackathon and second place at the 2017 Valtech Hackathon. The harder problem was timing: food transparency wasn't yet a mainstream concern, though initiatives such as Nutri-Score arrived soon after to answer exactly that.