What problem does it solve? Building production ML systems requires coordinating many disconnected stages—data validation, feature engineering, training, evaluation, and deployment—which often leads to fragile, non-reproducible workflows. This Skill provides structured guidance for orchestrating the full ML lifecycle into a coherent, automated pipeline. ## Core Features & Use Cases - Pipeline Architecture: Design DAG-based workflows with Airflow, Dagster, Kubeflow, or Prefect, including dependency management, retries, and error handling. - Training & Validation: Orchestrate training jobs, track experiments with MLflow or Weights & Biases, and run validation suites with regression detection. - Deployment Automation: Apply canary, blue-green, and shadow deployment patterns with rollback mechanisms and monitoring. - Use Case: A data science team needs to automate retraining of a churn model whenever new data arrives. Use this Skill to design a pipeline that validates incoming data, retrains the model, compares it against the baseline, and rolls it out gradually with canary deployment. ## Quick Start Ask the AI to design an end-to-end ML pipeline with data validation, training, evaluation, and canary deployment stages for your model.
