TL;DR — ML Pipeline Orchestrator

Design an ML pipeline orchestrator similar to Kubeflow Pipelines or MLflow that manages end-to-end machine learning workflows. The system defines pipelines as DAGs of steps (data ingestion, feature engineering, training, evaluation, deployment) and executes them reliably at scale. Key features: Define ML pipelines as DAGs with typed inputs/outputs. Execute pipelines with retry, timeout, and conditional logic.