Databricks Generative AI Engineer Associate Question 179
Select 3You are assembling a generative AI application in Databricks that leverages a fine-tuned LLM for summarizing customer support tickets. The application will be deployed as a REST API endpoint for use by multiple downstream systems. Which of the following steps are essential for successfully deploying and scaling this application on Databricks?
- A
Package the fine-tuned model and dependencies into a Databricks MLflow model.
- B
Expose the model as a REST API using Databricks Model Serving.
- C
Manually provision and configure additional compute resources for scalability.
- D
Enable request logging and monitoring using Databricks Model Serving metrics.
- E
Deploy the model to a Hadoop cluster for increased scalability.
Show answer and explanation
Correct answers: A, B, D
Explanation
To deploy and scale a generative AI application in Databricks, it is crucial to package the model using MLflow, expose it as a REST API with Databricks Model Serving, and enable monitoring with metrics. These steps ensure seamless integration, scalability, and observability within the Databricks ecosystem. Manual compute provisioning and deploying to Hadoop are not necessary due to Databricks' managed infrastructure and scalable serving capabilities.
- A. Correct.
Correct: Packaging the model and its dependencies into an MLflow model is an essential step for deployment and ensures reproducibility and easy integration with Databricks Model Serving.
- B. Correct.
Correct: Exposing the model as a REST API using Databricks Model Serving allows downstream systems to interact with the application efficiently.
- C. Incorrect.
Incorrect: Databricks automatically handles compute resource provisioning for Model Serving, so manual configuration is unnecessary.
- D. Correct.
Correct: Enabling request logging and monitoring using Databricks Model Serving metrics is essential for ensuring performance and identifying issues in production.
- E. Incorrect.
Incorrect: Deploying to a Hadoop cluster is not required for scalability, as Databricks Model Serving is designed to scale seamlessly within the Databricks environment.