tabular-ml-modeling

Features and purpose

State-of-the-art machine learning for large-scale tabular data (millions of rows, thousands of features). Use when building predictive models on tabular/structured data including regression and classification tasks. Covers GPU-accelerated gradient boosting (XGBoost, LightGBM, CatBoost), RAPIDS acceleration (cuDF, cuML), validation strategies, feature engineering, ensembling (hill climbing, stacking), and competition-winning techniques. Includes Numerai-specific optimizations with deep model parameters, era-based validation, multi-target ensembling, and payout scoring. Optimized for NVIDIA A100 GPU with CUDA 12.4.

Quick start

Follow the repository instructions for this skill and your agent. Review the files and dependencies before adding it to your workflow.