A Deployment Study of a Data-Driven Volunteer Engagement System for Food Security

We present and deploy an AI-driven volunteer engagement system for a large food rescue organization that combines optimization and machine learning to improve how volunteers are notified about food rescue opportunities. We develop both an optimized global notification policy and a rescue-specific recommender system with online planning, then validate the approach through real-world deployment and a randomized controlled trial. Our results demonstrate substantial improvements in volunteer engagement, rescue claim rates, and operational efficiency while reducing unnecessary notifications.