Build and tune ML model training

Build and train ML models with evaluation, hyperparameter tuning, and deployment readiness

End-to-end model development from data to evaluation when baselines and production requirements are specified

• · training data (features + labels)
• · problem type (classification/regression/clustering)
• · performance targets

• · trained model (pickled or format-agnostic)
• · evaluation metrics and cross-val results
• · hyperparameter sweep report
• · model card

• · scikit-learn/xgboost/pytorch
• · wandb/tensorboard
• · model serving framework

• · labeled data available
• · problem type clear
• · compute resources available

• · data imbalanced or too small
• · overfitting despite regularization
• · model not reproducible

• · uses cross-validation (not just train/test split)
• · logs hyperparameter sweeps
• · reports train/val/test metrics separately
• · model card documents assumptions