ML Regression Framework
Reusable ML regression framework standardizing feature engineering, training, validation, and scoring, adopted as the baseline for all subsequent platform models.
Akash Sharma
Primary Owner
Platform-Wide
Adoption
The Problem
Each new regression model required rebuilding the same boilerplate: feature pipeline, hyperparameter tuning, validation splits, scoring logic, and BigQuery writeback. This created code duplication and inconsistent engineering standards.
What Was Built
Built a configurable regression framework covering: feature group selection, outlier treatment, scaling, transformations; algorithm selection (XGBoost Regressor, Ridge); RandomizedSearchCV tuning; quarter-aware GroupShuffleSplit validation; MAPE/wMAPE/MAE/RMSE metrics; production scoring pipeline loading pickled models; BigQuery and Parquet writeback.
Business Impact
Reduced new model development time by standardizing the full ML lifecycle. Adopted as the baseline framework for all subsequent regression models on the platform.
Related Platform & Infrastructure Projects
Model Metric Dashboard
Internal ML governance dashboard tracking train/test MAPE, MAE, RMSE, and classification metrics across all deployed platform models.