MaxHalford/taxi-demo-rp-mz-rv-rd-st

Demonstrates end-to-end online learning with Materialize performing real-time feature engineering and performance monitoring via SQL, while River models continuously update from labeled data joined across pickup, dropoff, and prediction topics. Uses Redis as a model coordination layer between separate inference and learning services, enabling asynchronous model updates every 30 seconds without blocking predictions. The architecture showcases event-driven streaming with Redpanda as the central event log, where feature vectors are logged alongside predictions for later joining with ground-truth labels.

No commits in the last 6 months.