A universal churn prediction pipeline that automatically trains and selects the best model (e.g., logistic regression, random forest, gradient boosting) on any CSV with a “Churn” label and provides real‑time risk scoring via a Streamlit interface