Urban logistics is undergoing a transformation toward integrated, sustainable, and demand-responsive mobility solutions. This paper proposes a two-stage hierarchical clustering approach with parcel assignment (TSHCWAP) to address the computational challenges of optimizing large-scale ride-sharing services that combine passenger and parcel transportation under time window constraints. The proposed methodology applies agglomerative hierarchical clustering based on spatial trajectory distances to group transport requests for people, followed by the assignment of parcel requests to the nearest clusters. Strategic meeting points are introduced to improve operational efficiency by reducing the number of pickup and drop-off locations. The approach leverages the JAX computing framework to accelerate matrix operations, enabling the handling of high-volume, real-world datasets. Experimental evaluation using urban transport data from Madrid demonstrates a substantial reduction in computation time—up to 90%—with minimal impact on route quantity. Results confirm the model’s scalability and its potential for deployment in intelligent mobility systems seeking to balance service quality and resource efficiency.