Microsoft AI-300 Online Practice Questions and Exam Preparation

Microsoft AI-300 Online Questions & Answers

• Question 21:

  Note: This question is part of a series of questions that present the same scenario. Each question in the series contains a unique solution that might meet the stated goals. Some question sets might have more than one correct solution, while others might not have a correct solution.

  After you answer a question in this section, you will NOT be able to return to it. As a result, these questions will not appear on the review screen.

  You work in Microsoft Foundry with a prompt flow.

  You must manually evaluate prompts and compare results across prompt variants.

  You need to capture the inputs, outputs, token usage, and latencies for each flow run for the evaluation.

  Solution: Use the prompt flow SDK to enable tracing for the flow before executing runs. Then run the flow to generate traceable results.

  Does the solution meet the goal?

  A. Yes
  B. No
  B. No

  ---

  Explanation

  The solution is insufficient if it only stores or deploys prompt text without reviewable source-control history and evaluation. The requirement is not merely to keep a copy of prompts; it is to manage prompt lifecycle safely. Other approaches that skip Git-style versioning or variant comparison do not meet that operational rule.

• Question 22:

  HOTSPOT

  A machine learning model is deployed to production in Azure Machine Learning and is actively serving predictions for a business application.

  The model was trained by using a historical dataset that represented expected input patterns at the time of deployment.

  The team working on the model must ensure the following:

  1. Changes in input data distribution are detected.

  2. Appropriate actions are triggered when predefined thresholds are exceeded.

  You need to configure monitoring to meet the requirements.

  Which configuration should you use for each requirement? To answer, select the appropriate options in the answer area.

  NOTE: Each correct selection is worth one point.

  

  

  ---

  Map distribution-change detection to data drift monitoring and threshold-based action to alert or retraining triggers. Data drift is the production signal for detecting that incoming data no longer resembles the baseline data. Other actions, such as manual retraining or model registration alone, do not provide the monitoring and trigger behavior required by the scenario.

• Question 23:

  A Foundry workload processes regulated customer data. The workload must avoid public network exposure while still allowing platform services to reach required Azure resources.

  What should you configure?

  A. Private networking with managed identity and RBAC for dependent resources
  B. API keys embedded in each application configuration file
  C. Public endpoints restricted only by naming convention
  D. A larger prompt context window
  A. Private networking with managed identity and RBAC for dependent resources

  ---

  Explanation

  Private networking controls public exposure, while managed identity and RBAC give the Foundry workload controlled access to dependent Azure resources. Embedded API keys create secret-management risk, naming conventions are not a security boundary, and prompt context size does not address network isolation. The requirement is platform security, identity, and private connectivity.

• Question 24:

  A team deploys a model to a real-time endpoint in Azure Machine Learning. You deploy some updates to the endpoint.

  The endpoint returns errors after the new deployment is released.

  You need to restore the service as quickly as possible.

  What should you do first?

  A. Roll back traffic to the previous deployment.
  B. Delete the endpoint and immediately redeploy it.
  C. Change the authentication type to Azure Machine Learning token-based authentication.
  D. Increase the compute size.
  A. Roll back traffic to the previous deployment.

  ---

  Explanation

  Deployment logs are the first place to review when endpoint errors begin after an update because they expose startup failures, scoring exceptions, dependency issues, and deployment health events. Other responses such as retraining or changing traffic may be needed later, but the troubleshooting requirement starts with evidence from the endpoint deployment logs.

• Question 25:

  DRAG DROP

  A team deploys a generative AI application that uses a model deployed in Microsoft Foundry. The application must support latency monitoring under production load.

  You need to enable performance observability.

  Which three actions should you perform in sequence? To answer, move the appropriate actions from the list of actions to the answer area and arrange them in the correct order.

  Select and Place:

  

  

  ---

  The image answer enables monitoring metrics for the Foundry application. Latency monitoring under production load requires runtime performance observability such as latency, throughput, and response-time metrics. Other actions can improve release quality or model behavior, but they do not directly capture live performance signals for troubleshooting.

• Question 26:

  A production model shows declining accuracy because the distribution of incoming data has changed from the training data.

  Which monitoring signal should you investigate?

  A. Data drift
  B. Prompt fluency
  C. Git commit frequency
  D. Provisioned throughput utilization only
  A. Data drift

  ---

  Explanation

  Data drift is the monitoring signal for changes in production input distribution compared with the training baseline. Prompt fluency, Git activity, and throughput utilization measure other aspects of the system. The requirement describes accuracy decline from changed incoming data, which is exactly the data drift monitoring use case.

• Question 27:

  A company creates separate Microsoft Foundry projects for different product teams. Shared governance, connections, and platform settings must be centrally managed.

  Which structure best supports this model?

  A. A Foundry hub with separate project environments
  B. A separate Azure subscription for every prompt file
  C. A single notebook shared by all product teams
  D. A Machine Learning batch endpoint for each team
  A. A Foundry hub with separate project environments

  ---

  Explanation

  A Foundry hub with separate project environments supports centralized governance while letting teams work in isolated projects. Separate subscriptions or shared notebooks do not provide the same platform governance pattern, and batch endpoints are unrelated to organizing Foundry development. The requirement combines shared control with team-level isolation.

• Question 28:

  A prompt engineering team creates several prompt variants for the same application. The team must compare behavior and keep prompt history under source control.

  Which practice should you recommend?

  A. Store prompts and variants in a Git repository and evaluate them before release
  B. Store the active prompt only in a deployed endpoint name
  C. Register each prompt as an Azure Machine Learning compute target
  D. Disable prompt history after every successful deployment
  A. Store prompts and variants in a Git repository and evaluate them before release

  ---

  Explanation

  Storing prompts and variants in Git gives the team version history, review, and rollback, while evaluation before release compares behavior. Endpoint names, compute targets, and disabled history do not provide source control for prompt content. The requirement is prompt lifecycle management, not just runtime configuration.

• Question 29:

  HOTSPOT

  A team is building a generative AI agent by using Retrieval-Augmented Generation (RAG) in Microsoft Foundry.

  The team frequently updates prompt content. The team must be able to track changes across contributors while avoiding full application redeployments.

  You need to enable rapid prompt iteration with traceability. Applications consuming the agent must be able to use updated prompts without requiring redeployment.

  What should you configure for each requirement? To answer, select the appropriate options in the answer area.

  NOTE: Each correct selection is worth one point.

  

  

  ---

  Git gives the team reviewable history across prompt changes, and a CI/CD workflow can promote validated prompt revisions without ad hoc production edits. Other storage-only or environment-only choices do not provide both change tracking and controlled release behavior.

• Question 30:

  An Azure Machine Learning workspace contains multiple registered versions of a model that is used in production.

  An older model version must no longer be deployable, but it must remain available for compliance review and potential rollback.

  You need to change the state of the model version to meet the requirements.

  What should you do?

  A. Archive the training dataset for the model version.
  B. Delete the model version.
  C. Archive the model version.
  D. Unregister the model version.
  C. Archive the model version.

  ---

  Explanation

  Archiving the model version prevents normal deployment use while preserving the version for compliance review or rollback analysis. Deleting or unregistering the version would remove the record the organization needs to keep. Other asset actions do not change the lifecycle state of the specific registered model version.