CourseAI / ML Engineer: The Complete Course · Module 13: Specialisation · part 79 of 93
Part 79 · Module 13: Specialisation

Lesson 2: Depth in one track: tabular and risk

24 min read·
Module 13 · Lesson 2 · About 50 minutes · Needs: lesson 1, Module 3 lesson 3 (costs, calibration) and Module 6 lesson 2 (error analysis). The Kestrel data from course-guide/06.

The situation

Week 13, Thursday. Adaeze agreed: Kestrel's specialism is tabular and risk modelling. "Then show me what depth looks like," she says. "Not more models. What would a specialist know that we do not?"

One parcel runs through the lesson: KP0102652 (simulated), collected at Inverness on Tuesday 1 September 2026 for island route INV-R09, promised the next day. The ferries did not run that day. It was late, and the model gave it 0.041, too low for a warning.

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
Depth changes how Kestrel trains, gates and monitors the model and reads its late labels (blue); the rest stays as built (grey).

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