Skip to content

Exercises: Module 12 — Capstone Project Milestones

The capstone project uses this file for milestone checkpoints rather than traditional exercises. Each milestone produces a concrete, demonstrable deliverable.

Instructions

Complete each milestone in order. Check each item as you complete it. The milestones build on each other — do not skip ahead.

Milestone 1: Data Pipeline (Beginner–Intermediate)

Objective: Reproducible, documented data pipeline

  • Dataset selected and downloaded
  • Data validation script written (shape, missing values, data types, class balance)
  • Train/validation/test split created with stratification
  • Data statistics logged to MLflow
  • Script is runnable: python data_pipeline.py succeeds from scratch

Checkpoint: Show the MLflow UI with the data validation run.

Milestone 2: Baseline Model (Intermediate)

Objective: A working, tracked baseline

  • Classical ML baseline trained and evaluated with appropriate metrics
  • Experiment tracked in MLflow with parameters and metrics
  • Model registered in MLflow Model Registry as "staging"

Checkpoint: Show the MLflow experiment page with the baseline run.

Milestone 3: Improved Model (Intermediate–Advanced)

Objective: Hyperparameter search and model comparison

  • At least 10 different configurations trained and tracked
  • Best model identified and promoted to "production" in MLflow Registry

Checkpoint: Show a comparison table of all runs in MLflow.

Milestone 4: Serving API (Advanced)

Objective: A working, containerized prediction API

  • FastAPI /health and /predict endpoints implemented
  • Input validation with Pydantic
  • Prediction logging to SQLite
  • Model loaded from MLflow Registry
  • Docker container builds and runs: curl http://localhost:8000/health returns healthy

Checkpoint: Live API demonstration.

Milestone 5: Fairness Audit (Advanced)

Objective: Documented fairness analysis

  • Three fairness metrics computed
  • SHAP explanations for 10+ predictions
  • Governance memo written

Checkpoint: Run python fairness_audit.py. Show governance memo.

Milestone 6: Monitoring and Documentation (Expert)

Objective: Complete production-ready system with documentation

  • Drift detection script implemented and tested
  • Monitoring report written
  • Project README complete: setup, architecture, design decisions, limitations

Checkpoint: Run python monitoring.py. Show project README.

Final Submission Checklist

  • python data_pipeline.py && python train_model.py succeeds end-to-end
  • Docker API starts and responds to requests
  • MLflow shows 3+ experiment runs with best model as "production"
  • Fairness audit and governance memo complete
  • Monitoring script runs successfully
  • Project README explains the system to a new reader