Try 10 focused Microsoft AI-300 questions on GenAIOps infrastructure, with explanations, then continue with IT Mastery.
Open the matching IT Mastery practice page for timed mocks, topic drills, progress tracking, explanations, and full practice.
Try Microsoft AI-300 on Web View full Microsoft AI-300 practice page
| Field | Detail |
|---|---|
| Exam route | Microsoft AI-300 |
| Topic area | GenAIOps Infrastructure |
| Blueprint weight | 24% |
| Page purpose | Focused sample questions before returning to mixed practice |
Use this page to isolate GenAIOps infrastructure for Microsoft AI-300. Work through the 10 questions first, then review the explanations and return to mixed practice in IT Mastery.
| Pass | What to do | What to record |
|---|---|---|
| First attempt | Answer without checking the explanation first. | The fact, rule, calculation, or judgment point that controlled your answer. |
| Review | Read the explanation even when you were correct. | Why the best answer is stronger than the closest distractor. |
| Repair | Repeat only missed or uncertain items after a short break. | The pattern behind misses, not the answer letter. |
| Transfer | Return to mixed practice once the topic feels stable. | Whether the same skill holds up when the topic is no longer obvious. |
Blueprint context: 24% of the practice outline. A focused topic score can overstate readiness if you recognize the pattern too quickly, so use it as repair work before timed mixed sets.
These questions are original IT Mastery practice items aligned to this topic area. They are designed for self-assessment and are not official exam questions.
Topic: Design and Implement a GenAIOps Infrastructure
Your team created a Microsoft Foundry project environment for a generative AI service. The production resource uses private networking, deployments are automated, and operational evaluation workflows must run without a user’s interactive credentials. Which configuration should you verify to confirm the environment is ready for model deployment and evaluations?
Options:
A. A Git repository is connected and prompt files are versioned
B. A developer account can deploy a model from the portal
C. Project managed identity has RBAC, private network access, and evaluation tracing enabled
D. An Azure Machine Learning registry contains the model asset
Best answer: C
Explanation: A Foundry project environment is deployment-ready when its operational identity and network path support the full workflow: deploying or accessing the model, reading evaluation inputs, writing evaluation outputs, and emitting monitoring or tracing data. In a private-networked production setup, validating only an interactive portal action is insufficient because automation should use the project or workload managed identity. The same environment must also support evaluation workflow execution and observability, not just model hosting. Prompt source control is useful, but it does not prove the runtime can deploy and evaluate through the production network boundary.
Topic: Design and Implement a GenAIOps Infrastructure
A team is operationalizing a generative AI app in Microsoft Foundry. Platform engineers must prepare the Foundry resource, project environment, managed identity, RBAC assignments, and foundation-model deployment before developers release application changes. Developers should only consume the ready environment configuration at runtime.
Which configuration approach best meets this requirement?
Options:
A. Require each app release to manually configure the Foundry project
B. Provision the Foundry environment with Bicep and expose runtime settings
C. Store deployment names as constants in the application repository
D. Create Foundry resources from the application startup code
Best answer: B
Explanation: Separating environment readiness from application code changes means treating Foundry infrastructure and project configuration as platform-managed configuration, not as application logic. Bicep or Azure CLI can provision the Foundry resource, project environment, identity, RBAC, networking, and model deployment before the app is released. The application should then read environment-specific values such as endpoint names, deployment identifiers, or managed identity settings from configuration at runtime. This allows platform changes to be promoted, reviewed, and repeated without requiring developers to modify source code for each environment.
Topic: Design and Implement a GenAIOps Infrastructure
A team provisioned a Microsoft Foundry resource and project environment by using Bicep. The next GitHub Actions release will deploy a foundation model and run automated quality evaluations. The environment must keep public network access disabled, use managed identity, and store evaluation datasets in a private storage account. Which implementation should validate that the environment is ready?
Options:
A. Temporarily enable public access and test with a personal account.
B. Run a private-network smoke deployment and evaluation using managed identity.
C. Commit prompt variants to Git and defer deployment testing.
D. Confirm the Bicep deployment succeeded and approve the release.
Best answer: B
Explanation: Readiness validation for a Foundry project environment should test the operational path that production will use. A small smoke deployment plus a minimal automated evaluation confirms that the project can deploy a model, authenticate through managed identity, access private evaluation data, and write evaluation metrics, logs, or traces without relaxing network controls. This is stronger than checking provisioning status because IaC success does not prove runtime access or evaluation workflow connectivity. The key is to validate model deployment and evaluation execution under the same identity and private-network constraints that the release pipeline must preserve.
Topic: Design and Implement a GenAIOps Infrastructure
A team uses Microsoft Foundry prompt variants stored in Git for a customer-support summarization flow. The release policy says to promote the evaluated prompt version that passes both quality gates and has the best task fit.
| Variant | Groundedness gate: \(\ge 4.0\) | Safety finding rate gate: \(\le 1\%\) | Task-fit F1 |
|---|---|---|---|
prompt-v2 | 4.5 | 0.2% | 0.76 |
prompt-v3 | 4.4 | 0.6% | 0.81 |
prompt-v4 | 4.1 | 1.8% | 0.84 |
prompt-v5 | 3.8 | 0.3% | 0.86 |
Which implementation should the team use?
Options:
A. Promote prompt-v5 because it has the highest task-fit F1 overall.
B. Promote prompt-v2 because it has the lowest safety finding rate.
C. Promote prompt-v4 because it has the best passing groundedness with high F1.
D. Promote prompt-v3 and tag it as the approved Git prompt version.
Best answer: D
Explanation: Prompt variant comparison should apply release gates before choosing the best-performing candidate. In this scenario, the team must first exclude variants that fail groundedness or safety requirements, then compare task-fit F1 only among the remaining variants. prompt-v4 fails the safety gate, and prompt-v5 fails the groundedness gate. Both prompt-v2 and prompt-v3 pass, but prompt-v3 has the higher task-fit F1. Tagging the selected prompt in Git preserves the operational constraint for version-controlled prompt management and repeatable deployment.
prompt-v5 does not meet the groundedness gate.prompt-v4 exceeds the allowed safety finding rate.Topic: Design and Implement a GenAIOps Infrastructure
A GenAIOps team must provision identical Microsoft Foundry environments for dev, test, and prod. Each environment needs the same Foundry resources, managed identity, RBAC assignments, private networking settings, and foundation-model deployment, but with environment-specific names and capacity values. The provisioning process must be repeatable and reviewed through Git. Which implementation best meets these requirements?
Options:
A. Use a notebook to call Foundry APIs for each environment and save outputs as artifacts.
B. Create parameterized Bicep templates and deploy them with Azure CLI using per-environment parameter files.
C. Store prompt variants in Git and manually create Foundry resources before each release.
D. Create the dev environment in the portal, clone it manually, and document the differences.
Best answer: B
Explanation: Repeatable Foundry infrastructure provisioning should use infrastructure as code. Bicep templates define the desired Azure and Foundry-related resources declaratively, such as identities, RBAC assignments, networking configuration, and model deployments. Parameter files let the same template deploy dev, test, and prod with controlled differences, such as names, capacity, or endpoint settings. Azure CLI can run the deployment from an automation workflow, making the changes reviewable and reproducible from Git.
Manual cloning or notebook-driven provisioning can work for one-off setup, but they do not provide the same auditable, repeatable environment baseline.
Topic: Design and Implement a GenAIOps Infrastructure
A team uses Microsoft Foundry for a GenAIOps project named claims-assistant. A GitHub Actions workflow authenticates as a user-assigned managed identity and must update prompt variants and run evaluation jobs only in that project. The identity must not be able to change RBAC assignments, private networking, or assets in other Foundry projects. Which implementation should you use?
Options:
A. Assign Owner at the Foundry resource scope.
B. Assign Azure AI Developer at the claims-assistant project scope.
C. Assign Contributor at the resource group scope.
D. Assign Azure AI User at the claims-assistant project scope.
Best answer: B
Explanation: For Foundry GenAIOps work, RBAC should be assigned at the narrowest scope that supports the operation. Updating prompt variants and running evaluations are project-level asset activities, so the managed identity should receive a project-level role that can create and manage those assets. Azure AI Developer at the specific project scope fits that requirement while avoiding permissions to manage RBAC, private networking, or other projects. Broader Azure roles such as Owner or Contributor should be avoided for routine automation because they allow infrastructure or access changes outside the project boundary. Azure AI User is better suited to consuming or using assets, not managing project assets in an operations workflow.
Topic: Design and Implement a GenAIOps Infrastructure
A team evaluates a foundation model in Microsoft Foundry and approves it for a production GenAI app. After rollout, monitoring shows lower groundedness and higher response time than the approved run.
Deployment notes
| Item | Approved evaluation | Production deployment |
|---|---|---|
| Model | contoso-chat | contoso-chat |
| Model version | 2026-02-15 | latest |
| Prompt variant | support-v12 | support-v12 |
| Rollout | candidate only | 100% production |
What is the best root cause indicated by the evidence?
Options:
A. The endpoint needs provisioned throughput units.
B. The prompt variant was changed without source control.
C. The deployment did not pin the approved model version.
D. The team selected the wrong model family during evaluation.
Best answer: C
Explanation: This evidence points to a model versioning and deployment-control problem, not a model selection problem. The team evaluated contoso-chat version 2026-02-15, but production was configured to use latest. In a production GenAIOps workflow, the evaluated model version should be explicitly pinned or promoted through a controlled deployment strategy so monitoring results can be compared to the approved baseline. A 100% rollout amplifies the impact, but the mismatch comes from not deploying the exact evaluated version. Model selection criteria decide which model family and capabilities are appropriate; versioning controls which approved model release is actually used.
Topic: Design and Implement a GenAIOps Infrastructure
A team is operationalizing a Microsoft Foundry chat summarization feature for an internal support portal. Traffic is intermittent, the team does not need custom model weights or a dedicated GPU deployment, and operations wants provider-managed scaling with usage-based API access. Which implementation best preserves these constraints?
Options:
A. Deploy the model to a managed GPU compute endpoint.
B. Package the model as an Azure Machine Learning real-time endpoint.
C. Deploy the foundation model as a serverless API endpoint in Foundry.
D. Purchase provisioned throughput units for the model deployment.
Best answer: C
Explanation: Serverless API endpoints in Microsoft Foundry are appropriate when a foundation-model workload needs simple API access, provider-managed scaling, and does not require dedicated serving infrastructure. In this scenario, the workload is intermittent and the team explicitly wants to avoid managing GPU capacity or custom model-serving infrastructure. That makes a serverless deployment the best operational fit.
Dedicated managed compute or provisioned throughput is more appropriate when predictable capacity, isolation, or throughput planning is required. Azure Machine Learning real-time endpoints are primarily for operationalizing ML models rather than choosing the simplest Foundry foundation-model deployment path.
Topic: Design and Implement a GenAIOps Infrastructure
A GenAIOps team must provision the same Microsoft Foundry resource, project environment settings, managed identity, RBAC assignments, and private networking in development, test, and production subscriptions. The team wants repeatable deployments from source control with environment-specific values supplied at release time. Which configuration choice best meets the requirement?
Options:
A. Create each Foundry project manually in the portal
B. Use parameterized Bicep templates deployed by Azure CLI
C. Use prompt versioning to recreate project settings
D. Store setup steps in a shared runbook document
Best answer: B
Explanation: Repeatable Foundry infrastructure provisioning should use infrastructure as code. A Bicep template can define the Microsoft Foundry resource configuration, project environment settings, managed identities, RBAC assignments, and networking in a source-controlled, reviewable form. Parameters let the same template deploy to development, test, and production with different names, subscriptions, or network values while keeping the intended configuration consistent.
Manual portal setup and runbooks can describe the process, but they are more prone to drift and are harder to validate in pull requests. Prompt versioning is useful for tracking prompt assets, not provisioning Foundry infrastructure.
Topic: Design and Implement a GenAIOps Infrastructure
A Microsoft Foundry project stores prompt files in a Git repository. A production evaluation reports a prompt-version mismatch after a hotfix.
Exhibit:
run: prod-eval-042
reported_prompt: support-summary:v1.8
repo:
tag_v1.8: commit 8a41c2
main_HEAD: commit d9f07b
deployment_workflow:
checkout_ref: refs/heads/main
prompt_path: prompts/support-summary.md
recorded_prompt_version: v1.8
symptom: responses include a disclaimer added only in commit d9f07b
What is the best root cause?
Options:
A. The foundation model version changed during evaluation.
B. The evaluation dataset is mapped to the wrong output field.
C. The Git tag v1.8 was moved after deployment.
D. The workflow deployed from main while recording v1.8.
Best answer: D
Explanation: Git-based prompt version control should make each production prompt traceable to the exact repository state used for deployment. In this case, the workflow checks out the mutable main branch, but records support-summary:v1.8. The symptom confirms the endpoint used content from commit d9f07b, while the v1.8 tag points to commit 8a41c2. The diagnostic finding is a version-recording gap: the deployment process is not pinning or recording the immutable Git commit that supplied the prompt. A safer workflow records the commit SHA, tag, or prompt artifact version that was actually deployed.
tag_v1.8 still points to commit 8a41c2.d9f07b.Use the Microsoft AI-300 Practice Test page for the full IT Mastery practice bank, mixed-topic practice, timed mock exams, explanations, and web/mobile app access.
Try Microsoft AI-300 on Web View Microsoft AI-300 Practice Test
Read the Microsoft AI-300 Cheat Sheet for compact concept review before returning to timed practice.