Microsoft AI-103 (Developing AI Apps and Agents on Azure) Practice Test 5

Question 40

A development group needs its AI workflow to connect a Python application to its intended Foundry project instead of accidentally calling another environment. What technical approach most directly enables that?

  1. Run evaluated model changes through CI/CD gates
  2. Use private connectivity and correct DNS resolution
  3. Separate deployment names from model names
  4. Use the project's scoped endpoint and identity

Explanation

Correct Answer

D:

For the stated requirement to connect a Python application to its intended Foundry project instead of accidentally calling another environment, a project-scoped endpoint and appropriate credential select the target project and permitted resources; hardcoded secrets add avoidable risk; this is the mechanism that the system must implement.

Incorrect Answers

A:

For a system whose objective is to prevent a newly deployed model or prompt from reaching users before quality and safety checks pass, a delivery pipeline can test application behavior, evaluations and deployment health before promotion; versioned artifacts support rollback; this does not directly meet the required outcome to connect a Python application to its intended Foundry project instead of accidentally calling another environment.

B:

For a system whose objective is to keep Foundry resource access off public paths for a network-restricted application, private endpoints route traffic over approved private networking; proper private DNS and routing are prerequisites for clients to reach the service; this does not directly meet the required outcome to connect a Python application to its intended Foundry project instead of accidentally calling another environment.

C:

For a system whose objective is to invoke a specific deployed model from application code when a catalog model has multiple deployments, a model's catalog identity is not necessarily the deployment identifier used by inference requests; deployment references need environment-aware configuration; this does not directly meet the required outcome to connect a Python application to its intended Foundry project instead of accidentally calling another environment.

Question 41

A pilot has a functional gap: Inference calls fail with a deployment-not-found error despite the model being visible in the catalog. How should the team address that gap?

  1. Read the actual deployment identifier from the target environment and update application configuration
  2. Validate the model-and-deployment-type availability matrix for the target region before deployment
  3. Declare the missing resource and authorization settings in the reviewed deployment template
  4. Inject the correct environment-specific project endpoint and use a permitted identity credential

Explanation

Correct Answer

A:

The fault signal is that inference calls fail with a deployment-not-found error despite the model being visible in the catalog; the targeted fix works because a model's catalog identity is not necessarily the deployment identifier used by inference requests; deployment references need environment-aware configuration, restoring the ability to invoke a specific deployed model from application code when a catalog model has multiple deployments.

Incorrect Answers

B:

If the failing function were to avoid selecting a model or deployment type unsupported in the intended Azure region, this change could be relevant because Foundry model availability varies by model, region and deployment type; a catalog listing alone does not mean every combination can be deployed; the actual symptom is that inference calls fail with a deployment-not-found error despite the model being visible in the catalog, which calls for a different repair.

C:

If the failing function were to reproduce Foundry resources, roles, and deployment settings across test and production environments, this change could be relevant because infrastructure-as-code records resource configuration for reviewable repeatable deployments and reduces portal-only drift; secrets remain in secure stores; the actual symptom is that inference calls fail with a deployment-not-found error despite the model being visible in the catalog, which calls for a different repair.

D:

If the failing function were to connect a Python application to its intended Foundry project instead of accidentally calling another environment, this change could be relevant because a project-scoped endpoint and appropriate credential select the target project and permitted resources; hardcoded secrets add avoidable risk; the actual symptom is that inference calls fail with a deployment-not-found error despite the model being visible in the catalog, which calls for a different repair.

Question 42

An Azure AI workload passes basic health probes but has this issue: A workload with steady peak demand has variable latency, and compliance requires geography-limited inference. Which fix targets it?

  1. Select a supported Data Zone Standard deployment in the designated zone
  2. Choose a supported regional provisioned deployment and size PTUs using workload measurements
  3. Validate the model-and-deployment-type availability matrix for the target region before deployment
  4. Submit eligible workloads to a batch deployment and consume completed output asynchronously

Explanation

Correct Answer

B:

The fault signal is that a workload with steady peak demand has variable latency, and compliance requires geography-limited inference; the targeted fix works because regional provisioned deployments reserve model throughput within an Azure geography where supported; this couples residency with more predictable capacity, restoring the ability to support high predictable usage while requiring processing within an Azure geography.

Incorrect Answers

A:

If the failing function were to process prompts only within an approved Microsoft-defined data zone while demand fluctuates, this change could be relevant because data Zone Standard keeps inference processing within its specified zone while retaining pay-as-you-go characteristics; geography-specific processing is a different constraint; the actual symptom is that a workload with steady peak demand has variable latency, and compliance requires geography-limited inference, which calls for a different repair.

C:

If the failing function were to avoid selecting a model or deployment type unsupported in the intended Azure region, this change could be relevant because Foundry model availability varies by model, region and deployment type; a catalog listing alone does not mean every combination can be deployed; the actual symptom is that a workload with steady peak demand has variable latency, and compliance requires geography-limited inference, which calls for a different repair.

D:

If the failing function were to summarize a nightly backlog of records without any requirement for immediate responses, this change could be relevant because batch deployments are designed for large asynchronous jobs and favorable economics; they are not a synchronous low-latency chat endpoint; the actual symptom is that a workload with steady peak demand has variable latency, and compliance requires geography-limited inference, which calls for a different repair.

Question 43

After implementing a workflow to use private connectivity and correct DNS resolution, the team needs a robust release gate. Which check should it choose?

  1. Measure error rates and evaluated answer quality by deployment cohort during canary rollout
  2. Run drift detection and compare provisioned configuration with approved definitions
  3. Check provisioning preflight results for the required region and deployment type
  4. Resolve resource hostnames from the workload subnet and trace the resulting private network path

Explanation

Correct Answer

D:

The accepted change is to use private connectivity and correct DNS resolution; this evaluation checks whether it can keep Foundry resource access off public paths for a network-restricted application, since private endpoints route traffic over approved private networking; proper private DNS and routing are prerequisites for clients to reach the service.

Incorrect Answers

A:

This test concerns the ability to move a subset of application traffic to a newly validated model without changing every client, for which an application-level rollout control separates client configuration from backend deployment selection, enabling staged exposure and rollback; it does not establish whether the change to use private connectivity and correct DNS resolution works for the current release.

B:

This test concerns the ability to reproduce Foundry resources, roles, and deployment settings across test and production environments, for which infrastructure-as-code records resource configuration for reviewable repeatable deployments and reduces portal-only drift; secrets remain in secure stores; it does not establish whether the change to use private connectivity and correct DNS resolution works for the current release.

C:

This test concerns the ability to avoid selecting a model or deployment type unsupported in the intended Azure region, for which Foundry model availability varies by model, region and deployment type; a catalog listing alone does not mean every combination can be deployed; it does not establish whether the change to use private connectivity and correct DNS resolution works for the current release.

Question 44

Stakeholders want a production AI solution able to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions. Which choice implements that requirement?

  1. Run evaluated model changes through CI/CD gates
  2. Use regional provisioned throughput for stable localized capacity
  3. Use deployment aliases or versioned configuration for controlled rollout
  4. Use Global Standard for flexible general inference
  5. Use a Data Zone Standard deployment for zone boundaries

Explanation

Correct Answer

D:

For the stated requirement to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions, global Standard offers pay-per-token inference with flexible global routing; it does not reserve a fixed amount of dedicated throughput; this is the mechanism that the system must implement.

Incorrect Answers

A:

For a system whose objective is to prevent a newly deployed model or prompt from reaching users before quality and safety checks pass, a delivery pipeline can test application behavior, evaluations and deployment health before promotion; versioned artifacts support rollback; this does not directly meet the required outcome to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions.

B:

For a system whose objective is to support high predictable usage while requiring processing within an Azure geography, regional provisioned deployments reserve model throughput within an Azure geography where supported; this couples residency with more predictable capacity; this does not directly meet the required outcome to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions.

C:

For a system whose objective is to move a subset of application traffic to a newly validated model without changing every client, an application-level rollout control separates client configuration from backend deployment selection, enabling staged exposure and rollback; this does not directly meet the required outcome to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions.

E:

For a system whose objective is to process prompts only within an approved Microsoft-defined data zone while demand fluctuates, data Zone Standard keeps inference processing within its specified zone while retaining pay-as-you-go characteristics; geography-specific processing is a different constraint; this does not directly meet the required outcome to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions.

Question 45

The acceptance criteria for a customer-facing AI service require it to avoid storing long-lived service keys in source control while authorizing access to Foundry resources. Which design should be approved?

  1. Use managed identity for a production Azure-hosted app
  2. Use deployment aliases or versioned configuration for controlled rollout
  3. Promote infrastructure with declarative configuration
  4. Check model and region availability before provisioning

Explanation

Correct Answer

A:

For the stated requirement to avoid storing long-lived service keys in source control while authorizing access to Foundry resources, managed identity provides an Azure-hosted workload an Entra identity; it still needs the correct data-plane roles and reachable resource endpoint; this is the mechanism that the system must implement.

Incorrect Answers

B:

For a system whose objective is to move a subset of application traffic to a newly validated model without changing every client, an application-level rollout control separates client configuration from backend deployment selection, enabling staged exposure and rollback; this does not directly meet the required outcome to avoid storing long-lived service keys in source control while authorizing access to Foundry resources.

C:

For a system whose objective is to reproduce Foundry resources, roles, and deployment settings across test and production environments, infrastructure-as-code records resource configuration for reviewable repeatable deployments and reduces portal-only drift; secrets remain in secure stores; this does not directly meet the required outcome to avoid storing long-lived service keys in source control while authorizing access to Foundry resources.

D:

For a system whose objective is to avoid selecting a model or deployment type unsupported in the intended Azure region, Foundry model availability varies by model, region and deployment type; a catalog listing alone does not mean every combination can be deployed; this does not directly meet the required outcome to avoid storing long-lived service keys in source control while authorizing access to Foundry resources.

Question 46

In an architecture review, the key requirement is to reproduce Foundry resources, roles, and deployment settings across test and production environments. Which solution directly satisfies it?

  1. Separate deployment names from model names
  2. Run evaluated model changes through CI/CD gates
  3. Promote infrastructure with declarative configuration
  4. Use private connectivity and correct DNS resolution

Explanation

Correct Answer

C:

For the stated requirement to reproduce Foundry resources, roles, and deployment settings across test and production environments, infrastructure-as-code records resource configuration for reviewable repeatable deployments and reduces portal-only drift; secrets remain in secure stores; this is the mechanism that the system must implement.

Incorrect Answers

A:

For a system whose objective is to invoke a specific deployed model from application code when a catalog model has multiple deployments, a model's catalog identity is not necessarily the deployment identifier used by inference requests; deployment references need environment-aware configuration; this does not directly meet the required outcome to reproduce Foundry resources, roles, and deployment settings across test and production environments.

B:

For a system whose objective is to prevent a newly deployed model or prompt from reaching users before quality and safety checks pass, a delivery pipeline can test application behavior, evaluations and deployment health before promotion; versioned artifacts support rollback; this does not directly meet the required outcome to reproduce Foundry resources, roles, and deployment settings across test and production environments.

D:

For a system whose objective is to keep Foundry resource access off public paths for a network-restricted application, private endpoints route traffic over approved private networking; proper private DNS and routing are prerequisites for clients to reach the service; this does not directly meet the required outcome to reproduce Foundry resources, roles, and deployment settings across test and production environments.

Question 47

An Azure AI workload passes basic health probes but has this issue: Production lacks a required role binding that existed only through a manual portal step in test. Which fix targets it?

  1. Submit eligible workloads to a batch deployment and consume completed output asynchronously
  2. Inject the correct environment-specific project endpoint and use a permitted identity credential
  3. Add representative quality and safety evaluations as required checks before production promotion
  4. Declare the missing resource and authorization settings in the reviewed deployment template

Explanation

Correct Answer

D:

The fault signal is that production lacks a required role binding that existed only through a manual portal step in test; the targeted fix works because infrastructure-as-code records resource configuration for reviewable repeatable deployments and reduces portal-only drift; secrets remain in secure stores, restoring the ability to reproduce Foundry resources, roles, and deployment settings across test and production environments.

Incorrect Answers

A:

If the failing function were to summarize a nightly backlog of records without any requirement for immediate responses, this change could be relevant because batch deployments are designed for large asynchronous jobs and favorable economics; they are not a synchronous low-latency chat endpoint; the actual symptom is that production lacks a required role binding that existed only through a manual portal step in test, which calls for a different repair.

B:

If the failing function were to connect a Python application to its intended Foundry project instead of accidentally calling another environment, this change could be relevant because a project-scoped endpoint and appropriate credential select the target project and permitted resources; hardcoded secrets add avoidable risk; the actual symptom is that production lacks a required role binding that existed only through a manual portal step in test, which calls for a different repair.

C:

If the failing function were to prevent a newly deployed model or prompt from reaching users before quality and safety checks pass, this change could be relevant because a delivery pipeline can test application behavior, evaluations and deployment health before promotion; versioned artifacts support rollback; the actual symptom is that production lacks a required role binding that existed only through a manual portal step in test, which calls for a different repair.

Question 48

A system now includes a component to use the project's scoped endpoint and identity. Which evaluation should accompany the rollout?

  1. Measure batch completion, failures, and total cost against the overnight processing window
  2. Run drift detection and compare provisioned configuration with approved definitions
  3. Smoke-test the target project and resource identifiers in each deployed environment
  4. Exercise the deployment endpoint from a smoke test and check resolved deployment metadata

Explanation

Correct Answer

C:

The accepted change is to use the project's scoped endpoint and identity; this evaluation checks whether it can connect a Python application to its intended Foundry project instead of accidentally calling another environment, since a project-scoped endpoint and appropriate credential select the target project and permitted resources; hardcoded secrets add avoidable risk.

Incorrect Answers

A:

This test concerns the ability to summarize a nightly backlog of records without any requirement for immediate responses, for which batch deployments are designed for large asynchronous jobs and favorable economics; they are not a synchronous low-latency chat endpoint; it does not establish whether the change to use the project's scoped endpoint and identity works for the current release.

B:

This test concerns the ability to reproduce Foundry resources, roles, and deployment settings across test and production environments, for which infrastructure-as-code records resource configuration for reviewable repeatable deployments and reduces portal-only drift; secrets remain in secure stores; it does not establish whether the change to use the project's scoped endpoint and identity works for the current release.

D:

This test concerns the ability to invoke a specific deployed model from application code when a catalog model has multiple deployments, for which a model's catalog identity is not necessarily the deployment identifier used by inference requests; deployment references need environment-aware configuration; it does not establish whether the change to use the project's scoped endpoint and identity works for the current release.

Question 49

The planned deployment includes a mechanism to use asynchronous batch inference for delay-tolerant processing. Which outcome should the QA team measure?

  1. Confirm token-based access succeeds and the app can operate without embedded keys
  2. Measure batch completion, failures, and total cost against the overnight processing window
  3. Compare hourly demand variance, token spend, and successful requests before and after the change
  4. Load-test p95 latency, provisioned utilization, and the configured geography boundary

Explanation

Correct Answer

B:

The accepted change is to use asynchronous batch inference for delay-tolerant processing; this evaluation checks whether it can summarize a nightly backlog of records without any requirement for immediate responses, since batch deployments are designed for large asynchronous jobs and favorable economics; they are not a synchronous low-latency chat endpoint.

Incorrect Answers

A:

This test concerns the ability to avoid storing long-lived service keys in source control while authorizing access to Foundry resources, for which managed identity provides an Azure-hosted workload an Entra identity; it still needs the correct data-plane roles and reachable resource endpoint; it does not establish whether the change to use asynchronous batch inference for delay-tolerant processing works for the current release.

C:

This test concerns the ability to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions, for which global Standard offers pay-per-token inference with flexible global routing; it does not reserve a fixed amount of dedicated throughput; it does not establish whether the change to use asynchronous batch inference for delay-tolerant processing works for the current release.

D:

This test concerns the ability to support high predictable usage while requiring processing within an Azure geography, for which regional provisioned deployments reserve model throughput within an Azure geography where supported; this couples residency with more predictable capacity; it does not establish whether the change to use asynchronous batch inference for delay-tolerant processing works for the current release.

Question 50

A production AI solution is being scoped for production. The non-negotiable task is to keep Foundry resource access off public paths for a network-restricted application. Which implementation fits?

  1. Use private connectivity and correct DNS resolution
  2. Use deployment aliases or versioned configuration for controlled rollout
  3. Use managed identity for a production Azure-hosted app
  4. Use asynchronous batch inference for delay-tolerant processing

Explanation

Correct Answer

A:

For the stated requirement to keep Foundry resource access off public paths for a network-restricted application, private endpoints route traffic over approved private networking; proper private DNS and routing are prerequisites for clients to reach the service; this is the mechanism that the system must implement.

Incorrect Answers

B:

For a system whose objective is to move a subset of application traffic to a newly validated model without changing every client, an application-level rollout control separates client configuration from backend deployment selection, enabling staged exposure and rollback; this does not directly meet the required outcome to keep Foundry resource access off public paths for a network-restricted application.

C:

For a system whose objective is to avoid storing long-lived service keys in source control while authorizing access to Foundry resources, managed identity provides an Azure-hosted workload an Entra identity; it still needs the correct data-plane roles and reachable resource endpoint; this does not directly meet the required outcome to keep Foundry resource access off public paths for a network-restricted application.

D:

For a system whose objective is to summarize a nightly backlog of records without any requirement for immediate responses, batch deployments are designed for large asynchronous jobs and favorable economics; they are not a synchronous low-latency chat endpoint; this does not directly meet the required outcome to keep Foundry resource access off public paths for a network-restricted application.

Developing AI Apps and Agents on Azure (AI-103) Practice Tests