Practicum 2: End-to-End ML System, from Ingestion to Serving
An end-to-end ML system: from raw data to production serving
Instructor to be announced
About the course
Working in small teams, students build a full ML system around a real dataset: ingestion, feature engineering, training, CI/CD, deployment, and monitoring. The practicum reinforces material from the MLOps and Reliable ML courses under conditions as close to production as possible. The outcome is a working service with a documented SLO and a postmortem of its first iteration.
What you'll learn
✓Build an end-to-end ML service as a team of 2-3 with clearly divided responsibilities
✓Document an SLO (latency, availability, accuracy) and run load testing
✓Write a postmortem of the first production iteration with root cause analysis
Key topics
•Team-based ML development: roles, Git workflow, code review
•Data ingestion and validation with Great Expectations / Pandera
•Feature engineering pipeline with dbt or Feast
•Training pipeline with experiment tracking (MLflow/W&B)
•Deploying a REST/gRPC service with versioning
•Load testing and latency profiling
•Drift monitoring and alerting in production
This description was generated automatically and has not yet been reviewed by an instructor — it's a draft for discussion.