Databricks Machine Learning Associate Question 402
Select 3A data science team is deploying a machine learning model into production using Databricks. To ensure the long-term reliability and maintainability of their machine learning operations (MLOps) strategy, which of the following practices should they adopt?
- A
Version control for datasets, models, and code to track changes over time
- B
Relying solely on ad-hoc scripts for model deployment
- C
Monitoring deployed models for performance drift and data quality issues
- D
Building a centralized repository to store metadata, experiment results, and lineage
- E
Skipping testing of machine learning pipelines to accelerate deployment
Show answer and explanation
Correct answers: A, C, D
Explanation
To establish a strong MLOps strategy, teams should focus on reproducibility, monitoring, and transparency in their machine learning workflows. Version control, monitoring, and centralized repositories are essential practices that support these goals, while ad-hoc scripts and skipping testing introduce risks and inefficiencies.
- A. Correct.
Version control is a best practice in MLOps as it ensures reproducibility and traceability of changes to datasets, models, and code. This allows teams to debug issues, audit processes, and collaborate more effectively.
- B. Incorrect.
Using ad-hoc scripts for deployment is not a best practice as it can lead to inconsistencies, errors, and challenges in scaling. Automated and standardized deployment pipelines should be used instead.
- C. Correct.
Monitoring deployed models is critical to identify performance drift (e.g., due to data distribution changes) and ensure the quality of input data. Without monitoring, models may degrade over time, impacting business outcomes.
- D. Correct.
A centralized repository for metadata, experiment results, and lineage enables tracking and managing experiments efficiently, which is key to ensuring reproducibility and transparency in the MLOps lifecycle.
- E. Incorrect.
Skipping testing of machine learning pipelines is not a best practice as it increases the risk of deploying faulty or unreliable models. Comprehensive testing ensures robustness and reliability in production.