Weather sensors publish readings to MQTT topics. A Flask backend subscribes, parses payloads, and stores recent values for the dashboard. I containerized the stack with Docker so broker, backend, and dashboard start consistently. The main engineering challenges were reliable reconnect logic, message schema consistency, and handling out-of-order events. This mirrors production data pipelines: ingest → normalize → serve — the same pattern used in analytics and monitoring systems.