Databricks Generative AI Engineer Associate Question 85
Select 2You are building a generative AI application on Databricks to assist with text summarization. Your team wants to deploy the application as an endpoint to serve predictions in real-time. Which of the following steps are essential for setting up a real-time inference endpoint on Databricks?
- A
Create a Databricks MLflow model and register it in the MLflow Model Registry.
- B
Write the application code to load the model and process input data in a Databricks notebook.
- C
Use Databricks Lakehouse to directly serve the model without creating an endpoint.
- D
Deploy the model to a Databricks Serving endpoint and configure it for real-time inference.
- E
Set up a workflow in Databricks Jobs to invoke the endpoint and schedule batch predictions.
Show answer and explanation
Correct answers: A, D
Explanation
To set up a real-time inference endpoint on Databricks, you need to register the model using MLflow to manage its lifecycle and deploy it to a Databricks Serving endpoint for real-time predictions. While other steps like writing application code or creating batch workflows may be part of the overall application development, they are not directly related to the setup of a real-time endpoint.
- A. Correct.
Correct. Registering the model using MLflow is a fundamental step in preparing it for deployment and ensuring version control and traceability.
- B. Incorrect.
Partially correct but not essential. While writing application code might be a preparatory step, it is not directly part of setting up a real-time endpoint.
- C. Incorrect.
Incorrect. Databricks Lakehouse is not used to directly serve real-time ML models; it is a platform for managing data and analytics workloads.
- D. Correct.
Correct. Deploying the model to a Databricks Serving endpoint is necessary to enable real-time predictions.
- E. Incorrect.
Incorrect. While workflows in Databricks Jobs are useful for batch processing, they are not relevant for real-time inference.