MetaPlate: Counterfactual-Guided RAG-LLM Tool for Personalized Food Recommendation and Hyperglycemia Prevention

MetaPlate is a novel counterfactual-guided, context-aware decision-support framework designed to provide personalized meal recommendations for preventing postprandial hyperglycemia. It utilizes multimodal data, including continuous glucose monitoring (CGM) readings and user meal inputs, alongside a machine learning model for glucose response prediction and a retrieval-augmented generation (RAG) layer powered by an LLM for generating interpretable dietary suggestions. The system demonstrates significant improvements in meal realism and actionable recommendations, underscoring the value of integrating domain expertise and structured constraints in LLM applications for health-related decision support.