Question 12

Explanation

Correct Answer:
MLflow

Explanation:

1. Why This Is Correct?

   • MLflow is Databricks' open-source platform for managing the entire ML lifecycle, including:

     • Experiment Tracking: Log parameters, metrics, and models (mlflow.start_run(), mlflow.log_metric()).

     • Model Registry: Version, stage, and deploy models (mlflow.register_model()).

     • Projects: Package and reproduce runs (mlflow.projects.run()).

     • Model Serving: Deploy as REST APIs (mlflow models serve).

2. Example Workflow:

   
   

   import mlflow
   # Track experiment
   with mlflow.start_run():
       mlflow.log_param("learning_rate", 0.01)
       model = train_model()
       mlflow.sklearn.log_model(model, "model")
   # Register model
   mlflow.register_model("runs:/<run_id>/model", "prod_model")

Why Other Options Are Incorrect?

• Databricks Jobs:

  • Schedules/automates scripts (e.g., ETL), but lacks ML-specific tracking.

• Databricks Runtime:

  • Provides pre-installed ML libraries (e.g., TensorFlow), not lifecycle tools.

• AutoML:

  • Automates model training, but doesn’t manage deployment or tracking.

Key Takeaway:

For end-to-end ML lifecycle (experiment → production), use MLflow. Integrate with Databricks Jobs for scheduling and Runtime for execution.

Pro Tip: Use MLflow Pipelines (GA in 2023) for standardized MLOps workflows.