CRISP-DM, problem framing, business impact, and the Python/Jupyter environment

North-star and guardrail metrics; confusion matrices as business tools; dashboard workshop

Statistical significance, power, actionability, regression, classification, and validation pipelines

Calibration, overfitting, transparency, XGBoost/LightGBM, and preparation for model auditing

LIME, SHAP, counterfactuals, feature importance, partial dependence, and interaction effects

Slicing, error attribution, bias discovery, group fairness, proxy variables, and Fairlearn

Failure taxonomies for ML and LLM pipelines; logs, prompts, labels, and conversation analysis

Embeddings, clustering, semantic patterns, LLM judges, root-cause dashboards

Counterfactuals, propensity scores, matching, common support, and synthetic controls

Uplift modeling, heterogeneous treatment effects, Qini curves, ROI, and persuadables

Drift, feedback loops, silent failure, fairness decay, governance, Evidently, and Streamlit

The complete lifecycle, project showcase, industry deployment patterns, and final clinic