CourseAI / ML Engineer: The Complete Course · Module 14: Responsible ML · part 84 of 93
Part 84 · Module 14: Responsible ML

Lesson 2: Governance: records, privacy, attacks and saying no

27 min read·
Module 14 · Lesson 2 · About 50 minutes · Needs: lesson 1, reproducing a model (Module 10 lesson 1), personal data handling (Module 5 lesson 3), and the Kestrel data (course-guide/06).

The situation

Week 14, Thursday. Adaeze will sign the model card on Monday. Hannah asks three things first: can we show, months later, why a customer was warned; can analysts study the data without seeing who sent what; and does the law say anything about this model? Rhys adds a fourth: could someone bend the model on purpose?

The lesson follows the 06:00 warning for KP0088193 (simulated, lesson 1's parcel) on island route INV-R09, probability 0.2312, and ends with a decision record: should v2 warn island customers at all?

WHERE THIS FITS The Kestrel system this lesson SHARED SERVICES Depot scans parcels, routes Ingestion nightly batch Validation Pandera checks Data lake Parquet, DuckDB Experiments MLflow runs Training LightGBM, Optuna Online store Feast, Redis Features point in time Gates vs the champion Registry staged versions Serving API FastAPI, ONNX Consumers vans, customers Retraining trigger, budget Alerting on call Monitoring drift, Evidently Late labels a day later Orchestrator Prefect flows Versions DVC, Git CI/CD GitHub Actions Containers Docker images
Governance wraps the gates, the registry, versions and the serving path: blue marks the parts this lesson touches, grey the rest.

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.

Buy this course · ₹399See what is inside

Or open everything with a plan