Automated Credit Risk Predictive Modeling Pipeline

Design a comprehensive Python-based machine learning pipeline for credit risk assessment that integrates multiple data sources. Implement advanced feature engineering techniques, develop ensemble learning models (Random Forest, Gradient Boosting), and create a modular system for continuous model retraining. The solution must handle imbalanced datasets, provide model interpretability, and generate detailed risk scoring with confidence intervals.