01
Register the Model
Register AutoML's best model to Unity Catalog or MLflow Model Registry for lifecycle management.
02
Deploy Model Serving
Deploy to a Databricks Model Serving endpoint, exposing it as a REST API for real-time inference.
03
Enable Inference Tables
Automatically log all requests/responses to a Delta table in Unity Catalog. This is your source of truth.
04
Create Lakehouse Monitor
Configure an 'Inference Profile'. Specify Prediction, Timestamp, and optional Ground Truth Label columns.
05
Monitor Metrics & Dashboards
Auto-compute drift metrics (Data/Prediction Drift). Visualize in Databricks AI/BI dashboards.
06
Set Up Alerts
Create SQL-based alerts. Notify via Slack/PagerDuty if accuracy drops or significant drift is detected.
Action
07
Automate Feedback Loop
Trigger webhook to start a Databricks Job for retraining. Closes the loop.
Restarts at Step 1