A Machine Learning Engineer needs to build a credit risk model using Databricks. Due to strict data governance, production data cannot be accessed from development or staging environments. To manage MLOps, the engineer uses a "deploy code" strategy with separate development, staging, and production environments mapped to different catalogs in Unity Catalog. The CI/CD pipeline automates environment transitions. What is the primary architectural component promoted from staging to production to generate the final, production-ready model in this scenario?