ML engineers, what does your production ML pipeline look like? Describe your stack for data ingestion/validation, orchestration (e.g., Airflow/Kubeflow), training/experiment tracking, feature store, serving, and monitoring/retraining.