This chapter explores how integrating predictive modeling with route optimization can enhance the performance of urban meal delivery systems. Three routing strategies—Greedy, Vehicle Routing Problem (VRP), and VRP enhanced with LSTM-based predictive rebalancing—were evaluated across varying temporal periods throughout the day. Results show that while VRP reduces delivery durations compared to heuristic routing, the hybrid VRP+LSTM model achieves additional efficiency gains by anticipating spatial–temporal demand fluctuations. These improvements translate into lower delivery times, and greater operational stability. Policy implications emphasize the need for open urban data infrastructures, AI-driven optimization frameworks, and adaptive governance models to support sustainable last-mile logistics. The study demonstrates that hybrid predictive–optimization frameworks can significantly advance intelligent and sustainable urban delivery networks.