Google Professional Machine Learning Engineer Question 431
Select 3Google Cloud PlatformYou are working on a machine learning project hosted on Google Cloud, and your organization has strict compliance requirements for tracking and auditing metadata for all ML models deployed in production. Which of the following steps should you take to ensure proper metadata tracking and auditing?
- A
Use Vertex AI Metadata to track lineage, parameters, and artifacts associated with ML workflows.
- B
Enable Cloud Audit Logs for all services interacting with your ML models.
- C
Store all metadata in a local database for faster retrieval and analysis.
- D
Integrate Vertex AI Pipelines with a centralized metadata store for consistent tracking of pipeline executions.
- E
Manually document all training runs and model deployments in a shared spreadsheet.
Show answer and explanation
Correct answers: A, B, D
Explanation
To ensure proper metadata tracking and auditing for ML models in production, it is essential to use tools and services that provide automated, scalable, and consistent metadata management. Vertex AI Metadata and Vertex AI Pipelines are purpose-built for managing metadata in Google Cloud ML workflows. Additionally, enabling Cloud Audit Logs ensures that all interactions with cloud services are monitored and recorded, meeting compliance and auditing requirements. Manual processes or local storage are not suitable for scalable and compliant workflows.
- A. Correct.
Vertex AI Metadata is specifically designed for tracking ML workflow metadata, such as lineage, parameters, and artifacts, ensuring compliance and auditability.
- B. Correct.
Cloud Audit Logs provide a detailed trail of interactions with Google Cloud services, which is critical for meeting compliance requirements and auditing purposes.
- C. Incorrect.
Storing metadata in a local database may result in inconsistent tracking and lack of integration with Google Cloud services, making it unsuitable for scalable and compliant workflows.
- D. Correct.
Integrating Vertex AI Pipelines with a centralized metadata store ensures consistent and automated tracking of all pipeline executions, which is necessary for auditing and reproducibility.
- E. Incorrect.
Manually documenting metadata in a spreadsheet is error-prone and unsuitable for scalable and compliant workflows, especially in production environments.