The core of ML: the workflow, the algorithms, honest evaluation, and the judgment to pick the right tool. This is where you stop being someone who calls models and become someone who can defend one in a review. Covers the entire classic-ML core โ and goes beyond it.
Everything above, on one real dataset you care about (churn, fraud, credit, intrusion): EDA โ leakage-safe features โ cross-validated model comparison (linear โ trees โ boosting) โ SHAP explanations โ a small Streamlit app, plus one Kaggle competition entry. Write the Decision Lens first โ the capstone is the phase's real exam; the quizzes only check the vocabulary.
Builds: BYO-2 regression ยท BYO-3 trees/forest ยท BYO-4 k-means/PCA