CourseAI / ML Engineer: The Complete Course · Module 15: Professional Practice · part 89 of 93
Part 89 · Module 15: Professional Practice

Lesson 2: Growth: read, reproduce, write up, publish

24 min read·
Module 15 · Lesson 2 · About 50 minutes · Needs: Module 13 lesson 2 (reading a paper in three passes), Module 6 lesson 2 (the experiment log), lesson 1 of this module, and the Kestrel data (course-guide/06).

The situation

Week 15, Friday. Adaeze has two requests. "The repository goes public next month, as your portfolio and ours. Anyone should be able to rerun it from one command." And: "Rhys is interviewing ML engineers. Write the design answer you would want to hear, on Kestrel."

One claim runs through the lesson. Zhang, Bengio, Hardt, Recht and Vinyals (ICLR 2017) wrote: "Deep neural networks easily fit random labels." You will read it, reproduce it on a laptop CPU, reproduce it again on Kestrel's parcels, write it up with what did not reproduce, and publish it.

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
This lesson makes the repository public and reproducible: versions, CI, containers and the experiment record (blue); the rest of the system is 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