databricks-model-serving

To deploy a machine learning model to a Databricks Model Serving endpoint, use synchronous deployment for classical ML models like sklearn and xgboost, or custom PyFunc models, exposing them via scalable REST API endpoints for real-time predictions.

Yes, you can deploy complex GenAI agents like ResponsesAgent and LangGraph to Databricks Model Serving. This supports asynchronous job-based deployment and integrates tools such as Unity Catalog Functions and Vector Search indexes for advanced querying.

The best way to query deployed ML models or GenAI agents on Databricks is by using SDKs, REST APIs, or MCP tools. This enables real-time predictions and responses directly from your integrated applications and workflows.

Yes, Databricks Model Serving supports asynchronous job-based deployment specifically for GenAI agents. This contrasts with the synchronous deployment method used for standard classical ML models and custom Python functions.

You can expose a customer churn prediction model as an API endpoint by using synchronous deployment to Databricks Model Serving. Once deployed, your CRM system can query this REST API endpoint to flag at-risk customers in real-time.