Built a real-time data pipeline for a fleet of 1,900+ locomotives, aggregating sensor streams into a readable and predictive dashboard for drivers and depot dispatchers. Implemented wear detection and automated alerts for component replacement. Data was streamed in real-time via MQTT and WebSocket, processed with FastAPI and Python, stored in PostgreSQL, and delivered to the dispatcher interface through a Node.js backend. Stack: MQTT, FastAPI, Python, PostgreSQL, Node.js, WebSocket