Data Engineering
Bay Area Delivery Optimization
Polyglot persistence architecture combining PostgreSQL analytics and Neo4j graph algorithms for last-mile delivery optimization.
The Problem
Context & Challenge
Food delivery platforms need efficient routing and hub placement. The challenge was optimizing last-mile logistics across the Bay Area's complex BART transit network.
The Approach
Architecture & Implementation
Built a polyglot persistence architecture: PostgreSQL for ETL and ridership analytics (12 months of data), Neo4j for graph modeling (100+ nodes, 2,400+ edges). Applied Dijkstra shortest path, Louvain community detection, and PageRank centrality algorithms.
The Results
Impact & Metrics
Identified 9 optimal BART pickup locations and 11 community-based delivery hubs. Louvain clustering revealed natural service regions for localized delivery operations.
Key Result
Identified 9 pickup locations + 11 community hubs using Dijkstra, Louvain, and PageRank
Technologies & Methods
PythonSQLPostgreSQLNeo4jCypherDijkstraLouvain ClusteringPageRankETL