PMLE Collaborating Within and Across Teams to Manage Data and Models Practice Test 2 — 15 Questions

Question 1

A company uses BigQuery as their data warehouse. They want to version datasets for ML experiments and be able to query snapshots at specific points in time. Which approach is most cost-effective and requires minimal operational overhead?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Create BigQuery table snapshots at key milestones.. BigQuery table snapshots capture the table state at a point in time and are stored as a lightweight copy that shares the base table's data, so they are cheap and require no export or external tooling. They let you query the dataset as of a specific milestone, which matches the versioning requirement with minimal operational overhead. This is the most cost-effective native approach.

Answer

Use BigQuery table clones to create copies of the data.

Answer

Export the table to Cloud Storage as Parquet and use DVC to version the files.

Answer

Use BigQuery time-travel (7-day window) to query historical data without snapshots.

Question 2

A team is using Vertex AI Feature Store with an online store for low-latency serving. They notice increasing latency during peak hours. The feature data is updated frequently and requires strong consistency. Which online store type should they use?

Accepted Answer

Bigtable online store. Bigtable online store is recommended for high-throughput, low-latency, and strong consistency requirements.

Answer

Optimized online store

Answer

Cloud Spanner online store

Answer

Firestore online store

Question 3

An ML team uses Vertex AI Workbench managed notebooks and wants to version their notebook code and collaborate using Git. Which THREE steps are required to set up Git integration? (Select 3)

Accepted Answer

Configure Git credentials (username/email) using git config.. Option A is correct because setting your identity with git config (for example, git config --global user.name and user.email) is required before you can commit and push changes from the managed notebook to a Git repository. Option C is correct because the notebook environment must have the git client available; if it is not preinstalled, you must install it (for example, via apt-get or conda) before any Git operations can run. Option E is correct because you must clone the target repository with git clone in a notebook cell to obtain a local working copy that you can edit, commit, and push. Option B is not required: Cloud Source Repositories mirroring is an optional Google Cloud feature and is not a prerequisite for using Git with GitHub or another remote in Workbench. Option D is not required: there is no 'Vertex AI Notebooks API for Git sync' that must be enabled; Git integration relies on the standard git client and network access, not a dedicated Git-sync API.

Answer

Create a Cloud Source Repositories mirror of the GitHub repo.

Answer

Enable the Vertex AI Notebooks API for Git sync.

Question 4

A company is implementing MLOps with Vertex AI. They need to ensure that only approved models can be deployed to production. Which TWO practices should they adopt?

Accepted Answer

Use Vertex AI Model Registry aliases (e.g., 'champion') to mark production-ready models. Option C is correct because Vertex AI Model Registry aliases (such as 'champion') provide a controlled pointer to a specific model version, so deployment pipelines can reference only the alias that has been explicitly promoted to production-ready status, preventing unapproved versions from being served. Option E is correct because Vertex AI Model Evaluation combined with a manual approval gate enforces a human review step before a model is promoted or deployed, ensuring that only models meeting evaluation criteria and receiving sign-off reach production. Option A is wrong because using a single endpoint without versioning removes the ability to distinguish approved model versions and roll back, which undermines controlled deployment. Option B is wrong because granting any Vertex AI User the ability to deploy models bypasses approval controls and violates least privilege. Option D is wrong because storing models in Cloud Storage without the Model Registry eliminates the metadata, versioning, and alias governance needed to track and restrict which models are approved for production.

Answer

Use a single endpoint for all models without versioning

Answer

Allow any user with Vertex AI User role to deploy models

Answer

Store all models in Cloud Storage without registry

Question 5

A team wants to monitor features in Vertex AI Feature Store for drift. Which TWO configurations are required?

Accepted Answer

Enable feature monitoring on the feature view. Option A is correct because drift detection in Vertex AI Feature Store is enabled at the feature view level by turning on feature monitoring, which computes statistics and drift metrics against a baseline. Option C is correct because, to actually receive drift notifications, you must configure an alerting channel (such as email, Pub/Sub, or PagerDuty) so the monitoring job can send alerts when drift thresholds are exceeded. Option B is not required because monitoring results are stored automatically in the associated BigQuery or Cloud Storage location, not via a manually configured sink. Option D is not required because Vertex AI schedules the monitoring job itself once feature monitoring is enabled. Option E is not required because drift detection uses built-in statistical comparisons, not a deployed model endpoint.

Answer

Set up a BigQuery sink for monitoring results

Answer

Create a Cloud Scheduler job to trigger monitoring

Answer

Deploy a monitoring model to an endpoint

Question 6

Which THREE are valid uses of Vertex AI Metadata?

Accepted Answer

Record the execution of a pipeline step. Vertex AI Metadata is a managed ML metadata store that records artifacts, executions, and their lineage relationships, so option A is correct because each pipeline step runs as an Execution resource whose inputs, outputs, and parameters can be logged for reproducibility and auditing. Option B is correct because datasets and models are Artifacts, and Metadata lets you record the training Execution that links a specific dataset artifact to the resulting model artifact, capturing provenance. Option C is correct because Metadata's lineage tracking APIs (e.g., via the Vertex AI SDK's get_execution, get_artifact, and lineage queries) let you traverse relationships to discover which datasets, executions, or artifacts are upstream of a given model. Option D is not a Metadata function; deploying a model to an endpoint is done through Vertex AI Endpoints (e.g., Model.deploy or the endpoints API), not the metadata store. Option E is not the intended use of Metadata as a primary store; hyperparameter values belong to Vertex AI Experiments (backed by the Experiments/MLMD tracking), and while Metadata can log parameters on executions, storing experiment hyperparameters is the role of Vertex AI Experiments rather than a valid standalone use of Metadata.

Answer

Deploy a model to an endpoint

Answer

Store hyperparameter values of an experiment run

Question 7

A company is implementing MLOps on Google Cloud and needs to manage model versions, assign aliases (e.g., 'champion' for production, 'challenger' for staging), store evaluation metrics alongside each model version, and deploy models to endpoints. Which service should they use? (Choose THREE that are part of the solution.)

Accepted Answer

Vertex AI Model Registry. Vertex AI Model Registry (A) is correct because it is the service that manages model versions, lets you assign aliases such as 'champion' and 'challenger' to specific versions, and stores evaluation metrics alongside each version. Vertex AI Endpoints (C) is correct because deploying a model version to an endpoint is how the registered model is served for online predictions, and Model Registry integrates directly with endpoints for deployment. Vertex AI Pipelines (D) is correct because it orchestrates the MLOps workflow that registers models, assigns aliases, and deploys them, using components like ModelUploadOp and EndpointCreateOp to automate the process. Vertex AI Feature Store (B) is not part of this solution because it manages feature storage and serving for training and prediction, not model versioning, aliasing, or deployment. Vertex AI Experiments (E) is not part of this solution because it tracks and compares experiment runs and metrics during training, rather than managing model versions, aliases, or endpoint deployment.

Answer

Vertex AI Feature Store

Answer

Vertex AI Experiments

Question 8

A financial institution uses Vertex AI Pipelines to train models on sensitive customer data. They need to ensure that only authorized team members can view or modify pipeline runs and artifacts, while allowing data scientists to execute pipelines. Which combination of IAM roles and permissions should they assign?

Accepted Answer

Use the Vertex AI User role for data scientists and a custom role with only aiplatform.pipelineJobs.get and aiplatform.artifacts.get permissions for auditors.. The Vertex AI User role provides the necessary permissions to execute pipelines without administrative rights. A custom role with only get permissions on pipeline jobs and artifacts ensures auditors can view but not alter resources, adhering to least privilege. This pairing balances operational needs with strict access control for sensitive data.

Answer

Assign the Vertex AI Pipelines Operator role to data scientists and the Vertex AI Viewer role to auditors.

Answer

Grant the Vertex AI User role to data scientists and the Vertex AI Administrator role to auditors.

Answer

Grant the Vertex AI Service Agent role to data scientists and the Vertex AI User role to auditors.

Question 9

A data science team collaborates on a Vertex AI Workbench notebook. They need to share the notebook with another team member and allow them to edit and run cells. What is the simplest way to achieve this?

Accepted Answer

Use the notebook instance's sharing feature to add the team member as a user with edit permissions.. Vertex AI Workbench user-managed notebooks include a sharing feature that allows adding users with specific permissions. Granting edit permissions enables the team member to open the JupyterLab interface, modify notebooks, and run cells. This is the most direct and simple method, avoiding broader IAM changes or static exports.

Answer

Grant the other team member the Vertex AI User role on the project.

Answer

Export the notebook to a Cloud Storage bucket and send the link to the team member.

Answer

Share the notebook instance by granting the other team member the Notebooks Viewer role.

Question 10

An ML engineer needs to ensure that a model deployed to a Vertex AI endpoint uses the same feature preprocessing logic as during training. They have a preprocessing function written in Python. What should they do to achieve consistency?

Accepted Answer

Package the preprocessing function into a custom container and deploy it alongside the model.. A custom container packages the preprocessing function and the model together, ensuring the same code runs during training and serving. This eliminates training-serving skew by using identical transformations. Other options either do not apply preprocessing at serving time or introduce latency and inconsistency.

Answer

Deploy the model with a pre-built Vertex AI container and include the preprocessing function as a file in a Cloud Storage bucket.

Answer

Use Vertex AI Pipelines to preprocess data before sending it to the endpoint for prediction.

Answer

Use Vertex AI Feature Store to serve features and apply preprocessing at training time only.

Question 11

A machine learning team wants to share features across multiple models to reduce training-serving skew and ensure consistency. Which Vertex AI service should they use?

Accepted Answer

Vertex AI Feature Store. Vertex AI Feature Store centralizes feature storage, ensuring the same features are used for training and serving, reducing training-serving skew.

Answer

Vertex AI Workbench

Answer

Vertex AI Model Registry

Answer

Vertex AI Experiments

Question 12

An organization uses Vertex AI Pipelines and wants to track the lineage of datasets, models, and metrics across pipeline runs. They need to query upstream and downstream dependencies of an artifact. Which service should they use?

Accepted Answer

Vertex AI Metadata. Vertex AI Metadata stores the ML metadata graph produced by Vertex AI Pipelines, including artifacts (datasets, models, metrics), executions, and their input/output relationships. It exposes APIs to traverse this graph in both directions, so you can query upstream sources and downstream dependents of any artifact. This is the correct service for cross-run lineage queries.

Answer

Vertex AI Feature Store

Answer

Vertex AI Experiments

Answer

Vertex AI Model Registry

Question 13

A team uses Vertex AI Feature Store with an online store for low-latency serving. They need to support frequent updates to features (e.g., every minute) and require high write throughput (thousands of writes per second). Which online store type should they choose?

Accepted Answer

Bigtable online store. Bigtable online store is designed for high-throughput, low-latency serving with frequent updates, making it suitable for thousands of writes per second and minute-level feature refreshes. It leverages Bigtable's scalable NoSQL architecture, which handles high write loads efficiently. Optimized online store is for low-latency but may not sustain such high write throughput, while Firestore and Cloud SQL have lower write limits.

Answer

Optimized online store

Answer

Firestore online store

Answer

Cloud SQL online store

Question 14

A data engineer needs to version large datasets (multiple TB) in a Data Lake on Google Cloud. They require ACID transactions to ensure consistency when multiple jobs read/write concurrently. Which solution should they use?

Accepted Answer

Delta Lake on Dataproc. Delta Lake on Dataproc provides ACID transactions on cloud storage data lakes, enabling concurrent reads/writes with consistency.

Answer

BigQuery table snapshots

Answer

DVC (Data Version Control)

Answer

Vertex AI Feature Store

Question 15

A team wants to use Vertex AI Workbench for collaborative notebook development. They need a persistent environment that can be stopped and restarted without losing installed packages and data. Which instance type should they choose?

Accepted Answer

User-managed notebooks. User-managed notebooks are Compute Engine instances where the user controls the environment, including installed packages, kernel state, and persistent disk contents. Because the underlying VM and its boot/data disks persist across stop and start operations, installed packages and data remain intact. This makes them the right choice for a persistent, customizable environment.

Answer

Managed notebooks

Answer

Colab Enterprise notebooks

Answer

Vertex AI Pipelines