Quinn Desimone¹ and Julian Avellaneda², ¹USA, ²California State Polytechnic University, USA

Nutrition tracking tools often focus narrowly on macronutrients, and they completely neglect micronutritional accuracy, ingredient complexity, and personalization. Further, many applications also fail to address issues such as meal cost. This project addresses these gaps by developing an AI-powered adaptive meal recommendation application. This system integrates ingredient inventories, user health profiles, and predictive analytics to generate individualized meal plans. Being built with Flutter, Firebase, and large language models, the system provides real-time ingredient parsing, dietary restriction enforcement, and tailored nutrition guidance. This project was not simple, however. Several challenges were encountered during development. Thus includes inconsistent food classification, data synchronization issues across app pages, and ensuring the AI generated consistent outputs in the required JSON schema. These challenges were resolved through a myriad of methods. Including, backend integration, stricter prompting, and external database support. Two experiments were completed to evaluate blind spots: dietary restriction enforcement (mean accuracy 90.7) and nutritional information generation across food categories (perfect accuracy for meat, lower for complex foods). Results confirm strong baseline reliability but highlight the need for refinement in parsing compound ingredients and micronutrients. Ultimately, this system demonstrates the feasibility of delivering hyper-personalized, health-focused meal recommendations that advance beyond existing static or generalized solutions.

Health, Nutritional Info, Mobile application development via flutter, Inventory food management, Meal recommendation via health issues