ML Interview Roadmap

MLOps

Training Pipeline Orchestration

Cover scheduled and event-driven retraining, lineage, artifact storage, reproducibility, and failure recovery.

Recommended on day 3195 minutesIntermediate

Learning objectives

• Design batch, streaming, and triggered training pipelines
• Track lineage across data, features, code, parameters, and model artifacts
• Handle failed jobs, bad data, and rollback cleanly

Interview prompts

• What should be reproducible about a training run?
• When should retraining be scheduled versus event-driven?

Prerequisites

Data Validation and Data Quality Model Registry and CI/CD for ML