Build leak-safe sklearn model with calibration

You are asked to implement a complete scikit-learn pipeline that predicts whether a user will churn within the next 28 days (churn_28d). The prediction is made at a decision point snapshot_date, and the model may use only information that was known on or before that date. The dataset contains one record per user_id and snapshot_date, with these columns: user_id: string identifier snapshot_date: date representing the prediction point country: categorical field with about 180…