FoodTrust AI: The Food Safety Copilot

FoodTrust AI is an AI-powered food safety copilot built to help people trust what they eat before it reaches their body. For millions of people, food safety is deeply personal. A menu item marked “veg”, “eggless”, “halal”, “vegan”, or “safe” can still contain hidden risks like gelatin, egg wash, chicken stock, fish sauce, rennet, alcohol-based flavouring, shared fryer contamination, or vague terms such as “natural flavours”. For allergy sufferers, religious diners, vegetarians, parents, travellers, and international students, one unclear food description can create real physical, emotional, cultural, or religious harm. FoodTrust AI turns messy food information into a clear Safe / Ask First / Risky decision. Users can paste or upload a menu, receipt, ingredient label, delivery note, or restaurant description. They then select their personal profile, such as Jain, vegetarian, vegan, egg-free, dairy-free, halal, gluten-free, or specific allergens like peanut, tree nut, sesame, shellfish, fish, soy, milk, egg, wheat, and mustard. The system combines a deterministic dietary risk engine with Fireworks/Gemma AI enhancement. The rule engine catches known ingredient risks and cross-contact warnings, while the AI layer improves ingredient interpretation, explanation quality, and ambiguity detection. Instead of giving a blind yes/no answer, FoodTrust AI explains the evidence, highlights risk flags, and suggests what the user should ask before eating. FoodTrust AI also includes an incident mode for cases where food trust has already failed, helping users document what happened and prepare clearer reports for restaurants, delivery platforms, or support teams. Built with React, FastAPI, Docker, OCR/file parsing, Fireworks AI, and Gemma model integration, FoodTrust AI is designed as a practical safety layer for modern food decisions — because “probably safe” is not safe enough.