Skip to content

Foundations Track

This track builds core ML execution discipline: data understanding, feature preparation, robust splitting, and evaluation quality.

  1. projects/sota-supervised-learning-showcase
  2. projects/eda-leakage-profiling-showcase
  3. projects/feature-engineering-dimred-showcase
  4. projects/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.json
  • artifacts/eda/univariate_summary.csv
  • artifacts/eda/bivariate_vs_target.csv
  • artifacts/leakage/leakage_report.csv
  • artifacts/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?