Whenever you're ready, walk me through how you'd approach this.
Describe how you would define hard churn vs soft churn for Uber ride data, and how you would train a model to predict churn on a synthetic dataset.
Explain the observation and follow-up windows, label construction, treatment of censored riders, feature engineering from historical rides, class imbalance handling, temporal validation, threshold selection, and evaluation. Provide production-quality Python that generates or consumes synthetic rider-period data, trains a model, and reports appropriate metrics.