Foundations Track¶
This track builds core ML execution discipline: data understanding, feature preparation, robust splitting, and evaluation quality.
Recommended Sequence¶
projects/sota-supervised-learning-showcaseprojects/eda-leakage-profiling-showcaseprojects/feature-engineering-dimred-showcaseprojects/sota-unsupervised-semisup-showcase
Core Skills Covered¶
- Train/validation/test discipline.
- Stratified, group-aware, and time-aware split strategies.
- Univariate and bivariate EDA.
- Missingness diagnostics and leakage checks.
- Feature encoding, selection, and dimensionality reduction.
Evidence Artifacts To Inspect¶
artifacts/splits/split_manifest.jsonartifacts/eda/univariate_summary.csvartifacts/eda/bivariate_vs_target.csvartifacts/leakage/leakage_report.csvartifacts/selection/selection_scores.csv
Suggested Reflection Prompts¶
- Which split strategy is most defensible for this dataset and why?
- Which feature engineering step improved generalization most?
- Which leakage check would fail first if pipeline order changed?