Model Prototyping Using Notebooks (e.g., Gemini Enterprise Agent Platform Workbench and Colab Enterprise) PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A team has a working prototype in a Workbench notebook. They need to deploy the model to production, but the production environment has strict requirements for auditability, reproducibility, and automated retraining. The team wants to minimize the effort to move from prototype to production. What is the most effective strategy?

   • ACopy the notebook to a production Workbench instance and use a cron job to run it on a schedule, saving the model to Cloud Storage.
   • BUse the Vertex AI Prediction API to serve predictions from the notebook, and manually retrain the model when needed.
   • CUse Vertex AI Pipelines to package the notebook code into a reproducible training pipeline, and use Vertex AI Model Registry to manage model versions.
   • DDeploy the notebook directly to a Vertex AI Endpoint using the 'Deploy to endpoint' feature.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A large enterprise has multiple teams using Colab Enterprise. The security team wants to ensure that data scientists from different teams cannot access each other's notebooks. The teams are in different Google Cloud projects. What is the most effective way to isolate notebook access between teams?

   • ACreate a shared VPC and use network tags to isolate the notebook instances for each team.
   • BUse VPC Service Controls to create a separate perimeter for each team's project, preventing cross-project access.
   • CUse IAM conditions on the Colab Enterprise roles to restrict access to notebooks based on the project, and ensure each team's notebooks are in their own project.
   • DInstruct each team to use a different Google account and rely on the default Colab Enterprise access controls.
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A data science team uses Workbench instances for model prototyping. They want to implement a code review process where team members can comment on notebook code before changes are merged. The team also needs a full audit trail of who made changes and when. What is the most efficient way to meet these requirements?

   • AStore the notebooks in a Git repository, enable Git integration on the Workbench instances, and use the Git platform's pull request and commenting features for code review.
   • BExport the notebooks to Google Docs and use the suggesting and commenting features for review, then manually copy the final version back into Workbench.
   • CShare the notebook instances with the team and use the built-in commenting feature in Workbench to discuss changes, relying on the instance's activity log for the audit trail.
   • DCreate a shared notebook instance and instruct the team to make a backup copy of the notebook before making any changes, then review the backups manually.
   Reveal Solution Discussion Select an answer first
5. 9expert · hard

   A team is developing a scikit-learn model in a shared Colab Enterprise notebook. The notebook contains hard-coded credentials for a database. The team wants to refactor the code to remove the credentials and ensure they are not committed to the notebook's version history. What is the most secure way to handle the credentials?

   • AUse environment variables in the Workbench instance to store the credentials and access them in the notebook using os.environ.
   • BUse the Secret Manager API to store the credentials and retrieve them in the notebook using the google-cloud-secret-manager library, then delete the cell containing the hard-coded credentials.
   • CStore the credentials in a text file in the same Cloud Storage bucket as the notebook and read the file in the notebook.
   • DUse the Colab Enterprise 'Secrets' feature to store the credentials and access them with a special syntax in the notebook.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   An organization wants to prevent users from disabling VPC Service Controls on a Vertex AI Workbench environment. Which Google Cloud feature should be used to enforce this restriction?

   • ACloud DLP
   • BOrganization policy
   • CCloud Audit Logs
   • DIAM condition
   Reveal Solution Discussion Select an answer first