How Amtrak is building the data backbone for its largest transformation in over 50 years
Amtrak is modernizing its U.S. passenger rail network with new fleets and infrastructure while building a unified data platform on Databricks to integrate real-time telemetry, operational systems, and predictive analytics across its operations.
Amtrak is executing its largest physical transformation in over 50 years, introducing two new fleets—the NextGen Acela and Siemens Airo trainsets—while rebuilding aging infrastructure. To manage this scale, the railroad is developing Rail Intelligence, a digital platform on Databricks that consolidates fleet telemetry, reservation systems, capital project data, and wayside detector readings into a single governed environment. Previously, these systems operated in silos, limiting Amtrak’s ability to respond to equipment failures or connect fleet health with crew scheduling and passenger demand. The new platform aims to replace fragmented data workflows with a unified, real-time foundation for decision-making.
The platform aggregates data from over 100 sensors per trainset, dispatch systems, geospatial feeds, and the new Sqills S3 Passenger reservation system using Databricks’ Lakeflow Connect and real-time streaming. Raw data is processed through a medallion architecture in Delta Lake, standardized via Unity Catalog, and delivered as trusted data products with full lineage and quality controls. Machine learning models, managed through MLflow, power applications such as anomaly detection, defect analysis, and delay probability scoring. This infrastructure supports five intelligence products designed to improve operational outcomes across fleet management, safety, scheduling, reservations, and capital planning.
Fleet Health Intelligence provides predictive alerts for issues like door faults or bearing temperature deviations, enabling mechanical teams to address problems before failures occur. Safety Intelligence automates ride quality monitoring and uses refrigeration sensor data to identify food safety risks in café cars. Operational Readiness integrates fleet availability, crew scheduling, and maintenance windows into a real-time dashboard to optimize departures. Reservations Intelligence supports the transition from the legacy Arrow mainframe to the cloud-native Sqills S3 Passenger system, while Capital Prioritization uses live asset data to inform Amtrak’s $5.5 billion annual capital program.
Amtrak’s platform is currently live with ML anomaly detection running across multiple fleets, with plans to expand into fully predictive capabilities, including delay probability models and revenue prediction. Future enhancements include agentic workflows and natural language queries via Genie, as well as a Databricks Apps experience layer for self-service data access. The platform’s value grows as each new trainset, capital project, and passenger booking adds data, reinforcing the goal of a railroad that operates with greater reliability and efficiency through integrated, real-time intelligence.