A production-grade Databricks PySpark lakehouse on the Brazilian e-commerce (Olist) dataset, using a medallion architecture, three Spark MLlib models and MLflow tracking.