Food waste and inefficient distribution remain critical global challenges, with surplus food often failing to reach communities in need due to logistical constraints. This paper presents Replate, a multi-agent artificial intelligence system designed to optimize food redistribution by matching surplus providers with suitable NGOs. The system employs a coordinated agent-based architecture, where specialized agents perform data processing, constraint evaluation, and recommendation generation. Large Language Model (LLM)-based reasoning is integrated to enable contextual decision-making using factors such as location, capacity, and urgency. A multi-criteria scoring mechanism is used to rank potential matches and improve allocation efficiency. The system is implemented using LangGraph and FastAPI, enabling real-time interaction and deployment. Experimental results across multiple scenarios demonstrate improved matching relevance compared to baseline approaches, highlighting the effectiveness of combining multi-agent systems with LLM-driven reasoning for real-world resource optimization.
Rahul Pandey (Fri,) studied this question.