NYC Taxi Ride Time-Series Prediction

Ride demand spikes are hard to forecast without temporal feature engineering

Without a feature store, re-running pipelines recomputes all features from scratch

Hopsworks feature store caches time-series features; DagsHub + MLflow log every training run

Time-series models (lag features, rolling stats) updated on schedule via pipeline

Predictions visible on a live Streamlit app refreshed each run

All experiment runs reproducible via DagsHub MLflow tracking

Feature pipeline separated from training — swap models without reprocessing data