A Generative AI Engineer has already trained an LLM on Databricks and it is now ready to be deployed. Which of the following steps correctly outlines the easiest process for deploying a model on…

Question

Options

• ALog the model as a pickle object, upload the object to Unity Catalog Volume, register it to Unity
• BLog the model using MLflow during training, directly register the model to Unity Catalog using the
• CSave the model along with its dependencies in a local directory, build the Docker image, and run
• DWrap the LLM's prediction function into a Flask application and serve using Gunicorn

• A
  3% (1)
• B
  93% (27)
• D
  3% (1)

Explanation

The easiest and recommended Databricks-native workflow for deploying a trained LLM is to log the model with MLflow during training and then register it directly to Unity Catalog using the MLflow API (e.g., mlflow.register_model()). From there, the model can be deployed to a Databricks Model Serving endpoint with minimal additional steps. This approach leverages Databricks' built-in integrations and requires no extra infrastructure work. Option A (pickle + Unity Catalog Volume) is a non-standard, manual approach. Option C (Docker image build) adds significant complexity and is unnecessary when using Databricks' managed serving. Option D (Flask + Gunicorn) bypasses all of Databricks' model serving capabilities, requiring the engineer to manage their own web server infrastructure.

No community discussion yet for this question.