Microsoft AI-300 Online Practice Questions and Exam Preparation

Microsoft AI-300 Online Questions & Answers

• Question 61:

  A domain-specific generative AI solution performs poorly on specialized terminology even after retrieval tuning. The team plans to customize the model and must manage the customized model through production.

  Which activity is in scope?

  A. Advanced fine-tuning with performance monitoring of the fine-tuned model
  B. Deleting all evaluation datasets after deployment
  C. Replacing source control with manual prompt edits
  D. Disabling observability to reduce logging cost
  A. Advanced fine-tuning with performance monitoring of the fine-tuned model

  ---

  Explanation

  Advanced fine-tuning with ongoing performance monitoring is appropriate when retrieval tuning is not enough for specialized terminology. The customized model must still be managed through development and production. Deleting evaluation data, bypassing source control, or disabling observability weakens quality control and violates the operational requirement.

• Question 62:

  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: In Microsoft Foundry, turn on Tracing for the prompt flow of the project and execute test runs to produce trace data.

  Does the solution meet the goal?

  A. Yes
  B. No
  A. Yes

  ---

  Explanation

  The solution meets the goal when it stores prompts in source control and uses a controlled review path before release. That provides history, collaboration, and rollback for prompt assets. Other methods that edit production directly or store prompts without review cannot meet the requirement for governed prompt lifecycle management.

• Question 63:

  DRAG DROP

  You manage an Azure Machine Learning model lifecycle.

  You need to match each operational requirement in the exhibit to the best action.

  Select and place the correct action for each requirement.

  Select and Place:

  

  

  ---

  Requirement 1 = Configure MLflow experiment tracking;

  Requirement 2 = Register the MLflow model;

  Requirement 3 = Split traffic between deployments MLflow experiment tracking records metrics, parameters, and artifacts for training jobs. Registering an MLflow model makes a specific model artifact version available for lifecycle management and deployment.

  Splitting traffic between deployments is the controlled rollout mechanism for validating a new deployment against limited production traffic. Archiving a model version is a lifecycle action for preventing normal use of an older version, not for tracking, registration, or progressive rollout.

• Question 64:

  A production generative AI application needs troubleshooting data for failed requests. Operators must trace request flow and inspect detailed runtime behavior.

  What should you enable?

  A. Detailed logging, tracing, and debugging capabilities
  B. Only a larger chunk size for retrieval
  C. Only model archiving for old versions
  D. A manual screenshot of each user session
  A. Detailed logging, tracing, and debugging capabilities

  ---

  Explanation

  Detailed logging, tracing, and debugging provide the request-level evidence operators need to troubleshoot failed production generative AI calls. RAG chunk sizing, model archiving, and screenshots solve other problems and are not scalable observability mechanisms. The requirement is tracing runtime behavior, so observability instrumentation is the correct control.

• Question 65:

  You manage an Azure Machine learning workspace. You develop a machine learning model.

  You must deploy the model to use a low-priority VM with a pricing discount.

  You need to deploy the model.

  Which compute target should you use?

  A. Azure Container Instances (ACI)
  B. Azure Machine Learning compute clusters
  C. Local deployment
  D. Azure Kubernetes Service (AKS)
  B. Azure Machine Learning compute clusters

  ---

  Explanation

  Azure Machine Learning compute clusters support training jobs and can use low-priority virtual machines for discounted compute. Azure Container Instances and local deployment are not the standard managed training target for this requirement, and AKS is generally used for Kubernetes-based serving scenarios.

  The other choices do not combine managed training execution with low-priority VM economics.

• Question 66:

  HOTSPOT

  A team operates a production generative AI application in Microsoft Foundry. Users report intermittent slow responses during peak business hours.

  You need to identify the area in the exhibit that should be used for continuous runtime monitoring of the live application.

  To answer, select the appropriate options in the answer area.

  

  

  ---

  Continuous monitoring is the operational mechanism for observing a live generative AI application after deployment. The exhibit shows latency, throughput, and response time in Area B, which are the runtime signals needed for production troubleshooting. Offline evaluation, prompt versioning, RAG tuning, and cost review are useful practices, but they do not provide the primary runtime view for intermittent production slowdowns.

• Question 67:

  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 manage an Azure Machine Learning workspace. The Python script named script.py reads an argument named training_data. The training_data argument specifies the path to the training data in a file named dataset1.csv.

  You plan to run the script.py Python script as a command job that trains a machine learning model. You need to provide the command to pass the path for the dataset as a parameter value when you submit the script as a training job.

  Solution: python train.py --training_data training_data.

  Does the solution meet the goal?

  A. Yes
  B. No
  B. No

  ---

  Explanation

  The proposed solution meets the goal because MLflow tracking captures experiment metadata, metrics, parameters, and artifacts in a repeatable way. That provides the consistent experiment record needed for comparison. Other logging systems can be useful operationally, but they do not replace MLflow as the model-training experiment tracking mechanism in this requirement.

• Question 68:

  A financial services company is deploying Microsoft Foundry to host generative AI workloads that process regulated customer data.

  The Microsoft Foundry environment must prevent any public network exposure while still allowing services managed by Microsoft Foundry to communicate with dependent Azure resources.

  Security auditors require that all traffic to and from the Microsoft Foundry resource remain on private networks, with no public endpoints available.

  You need to configure the Microsoft Foundry environment so that network access is restricted while maintaining full platform functionality.

  Which two actions should you perform? (Choose two.)

  A. Configure a managed virtual network for the Microsoft Foundry resource.
  B. Use API key authentication for all model endpoints.
  C. Deploy the Microsoft Foundry resource in a separate Azure subscription.
  D. Disable public network access to the Microsoft Foundry resource.
  E. Disable all inbound network access.
  A. Configure a managed virtual network for the Microsoft Foundry resource.
  E. Disable all inbound network access.

  ---

  Explanation

  A managed virtual network and disabled public network access together keep the Foundry environment private while still allowing required managed service communication. API keys, separate subscriptions, or blocking all inbound communication do not satisfy the full platform requirement. The requirement is private networking with functional dependent-resource access, not simply credential or subscription separation.

• Question 69:

  DRAG DROP

  A team deploys a machine learning model to a managed online endpoint. The team monitors model performance and data quality metrics in production.

  When monitoring thresholds are exceeded, the team requires an automated operational response that notifies downstream systems.

  You need to configure the monitoring solution to meet the requirements.

  Which configuration should you associate with each requirement as a first step? To answer, move the appropriate configurations to the correct requirements. You may use each configuration once, more than once, or not at all. You may need to move the split bar between panes or scroll to view content.

  NOTE: Each correct selection is worth one point.

  Select and Place:

  

  

  ---

  The image separates monitoring, alerting, and retraining responsibilities. Alert rules provide the automated response when thresholds are crossed, while retraining pipelines address model quality degradation rather than infrastructure optimization. Other configurations such as autoscaling are useful for capacity, but they do not directly implement model performance monitoring and retraining requirements.

• Question 70:

  A company's platform engineers manage the resource settings and governance of Microsoft Foundry.

  Developers must be able to create and update project assets but must not be able to change resource-level configurations.

  You need to enforce least privilege access for the engineers and developers.

  Which two actions should you perform? (Choose two.)

  A. Assign a resource-level Azure AI Administrator role to the platform engineers.
  B. Disable Microsoft Entra ID authentication for the Microsoft Foundry resource.
  C. Assign the Azure AI Developer role to the developers.
  D. Share a single API key across all teams.
  A. Assign a resource-level Azure AI Administrator role to the platform engineers.
  C. Assign the Azure AI Developer role to the developers.

  ---

  Explanation

  Developers should receive a role that lets them create and update project assets, while platform engineers retain broader governance and resource-setting permissions. RBAC and managed identity configuration are the right controls for separating duties. Other choices that grant full ownership or rely on shared secrets would violate least privilege and the governance requirement.