user@devops:~$ cat README.md
ML Production Pipeline
# Description
ML Production Pipeline is a comprehensive simulation of an end-to-end machine learning system in production. It covers data validation with quality checks, custom feature engineering (tenure segments, charge ratios, service interactions), model training with 3 algorithms (LogisticRegression, RandomForest, GradientBoosting) using 5-fold stratified cross-validation, model registry with MLflow experiment tracking and versioned artifacts, batch inference across 5 production batches with simulated drift, and continuous monitoring using PSI (Population Stability Index) with accuracy/AUC tracking. Includes 8 visualizations covering full pipeline architecture, model comparison, ROC curves, confusion matrix, feature importance, and monitoring dashboards.
# Key features
$ Data validation and quality reporting with missing value detection
$ Custom feature engineering pipeline with tenure segmentation, charge ratios, and service interaction features
$ Scikit-learn ColumnTransformer for mixed numeric/categorical/binary preprocessing
$ 3-model training pipeline (LogisticRegression, RandomForest, GradientBoosting) with 5-fold cross-validation
$ MLflow experiment tracking with model registry and nested runs
$ Batch inference pipeline across 5 production batches with versioned model persistence
$ Population Stability Index (PSI) drift detection for monitoring score distribution shifts
$ Real-time monitoring dashboard with accuracy/AUC trends and PSI alert thresholds
# Gallery
# Technologies used
