FoodGuard: AI-Driven Personalized Food Safety Communication System

Moonshot Award

Product or Company Development

Abstract

Food safety risk is growing faster than the expert support available to manage it. Small and mid-sized manufacturers carry the same regulatory burden as national brands without the staff to match, and turnover drains the knowledge they have. General-purpose AI can only give generic answers non-specific to the manufacturing conditions. FoodGuard (now FLX agent platform) asks whether AI can widen access to that expertise without replacing the expert. We built a platform that draws its answers from a curated library of authoritative sources: FDA, USDA, industry standards, and peer-reviewed research. 

The system is built upon an ontology for food safety related concepts, so a question about one organism surfaces the sanitation, design, and testing evidence tied to it. A separate tool pulls current federal regulation text when a question turns on compliance. The platform runs structured workflows written by food safety scientists, not open-ended chat. It interviews a company about its facility, products, and processes, then builds a risk assessment or environmental monitoring program step by step. Each recommendation carries its source, states what remains uncertain, and routes high-risk questions to a human reviewer. 

A validation framework in development scores accuracy, consistency, usefulness, and agreement with experts.