Automated Credit Risk Scoring Model with Machine Learning

Develop a comprehensive credit risk assessment pipeline using Python that integrates multiple data sources including historical loan performance, credit bureau data, and alternative financial indicators. Implement a scikit-learn-based predictive model that can generate risk scores with 85%+ accuracy, utilizing ensemble methods like Random Forest and gradient boosting. The solution should include automated feature engineering, cross-validation strategies, and a Flask API endpoint for real-time scoring. Include robust error handling, logging, and a methodology for model drift detection and periodic retraining.