Question 60
A project needs to complete all of these distinct tasks: (1) process prompts only within an approved Microsoft-defined data zone while demand fluctuates; and (2) summarize a nightly backlog of records without any requirement for immediate responses. Select one suitable capability for each task. Choose TWO.
- Use asynchronous batch inference for delay-tolerant processing
- Use private connectivity and correct DNS resolution
- Separate deployment names from model names
- Use regional provisioned throughput for stable localized capacity
- Use a Data Zone Standard deployment for zone boundaries
- Use Global Standard for flexible general inference
Explanation
Correct Answers
A:For the stated requirement 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 is the mechanism that the system must implement. This choice satisfies the specific deliverable to summarize a nightly backlog of records without any requirement for immediate responses, while the other deliverables require separate capabilities.
E:For the stated requirement 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 is the mechanism that the system must implement. This choice satisfies the specific deliverable to process prompts only within an approved Microsoft-defined data zone while demand fluctuates, while the other deliverables require separate capabilities.
Incorrect Answers
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 process prompts only within an approved Microsoft-defined data zone while demand fluctuates.
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 process prompts only within an approved Microsoft-defined data zone while demand fluctuates.
D: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 process prompts only within an approved Microsoft-defined data zone while demand fluctuates.
F:For a system whose objective is 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 does not directly meet the required outcome to process prompts only within an approved Microsoft-defined data zone while demand fluctuates.
Question 61
What test should engineers run after they promote infrastructure with declarative configuration, rather than relying solely on a successful API response?
- Smoke-test the target project and resource identifiers in each deployed environment
- Check provisioning preflight results for the required region and deployment type
- Check pipeline rejection on a deliberately failing evaluation and verify rollback readiness
- Run drift detection and compare provisioned configuration with approved definitions
Explanation
Correct Answer
D:The accepted change is to promote infrastructure with declarative configuration; this evaluation checks whether it can reproduce Foundry resources, roles, and deployment settings across test and production environments, since infrastructure-as-code records resource configuration for reviewable repeatable deployments and reduces portal-only drift; secrets remain in secure stores.
Incorrect Answers
A:This test concerns the ability to connect a Python application to its intended Foundry project instead of accidentally calling another environment, for which a project-scoped endpoint and appropriate credential select the target project and permitted resources; hardcoded secrets add avoidable risk; it does not establish whether the change to promote infrastructure with declarative configuration works for the current release.
B: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 promote infrastructure with declarative configuration works for the current release.
C:This test concerns the ability to prevent a newly deployed model or prompt from reaching users before quality and safety checks pass, for which a delivery pipeline can test application behavior, evaluations and deployment health before promotion; versioned artifacts support rollback; it does not establish whether the change to promote infrastructure with declarative configuration works for the current release.
Question 62
A user-acceptance test fails with the following observation: A prototype reserves expensive capacity yet uses it sporadically, and no residency limit applies. What is the best correction?
- Switch the pilot to an eligible Global Standard deployment and monitor metered usage
- Choose a supported regional provisioned deployment and size PTUs using workload measurements
- Fix private DNS linkage and ensure the client network reaches the private endpoint
- Submit eligible workloads to a batch deployment and consume completed output asynchronously
Explanation
Correct Answer
A:The fault signal is that a prototype reserves expensive capacity yet uses it sporadically, and no residency limit applies; the targeted fix works because global Standard offers pay-per-token inference with flexible global routing; it does not reserve a fixed amount of dedicated throughput, restoring the ability to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions.
Incorrect Answers
B:If the failing function were to support high predictable usage while requiring processing within an Azure geography, this change could be relevant because regional provisioned deployments reserve model throughput within an Azure geography where supported; this couples residency with more predictable capacity; the actual symptom is that a prototype reserves expensive capacity yet uses it sporadically, and no residency limit applies, which calls for a different repair.
C:If the failing function were to keep Foundry resource access off public paths for a network-restricted application, this change could be relevant because private endpoints route traffic over approved private networking; proper private DNS and routing are prerequisites for clients to reach the service; the actual symptom is that a prototype reserves expensive capacity yet uses it sporadically, and no residency limit applies, 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 prototype reserves expensive capacity yet uses it sporadically, and no residency limit applies, which calls for a different repair.
Question 63
A customer-facing AI service is being scoped for production. The non-negotiable task is to avoid selecting a model or deployment type unsupported in the intended Azure region. Which implementation fits?
- Use regional provisioned throughput for stable localized capacity
- Use private connectivity and correct DNS resolution
- Use managed identity for a production Azure-hosted app
- Check model and region availability before provisioning
Explanation
Correct Answer
D:For the stated requirement 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 is the mechanism that the system must implement.
Incorrect Answers
A: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 avoid selecting a model or deployment type unsupported in the intended Azure region.
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 avoid selecting a model or deployment type unsupported in the intended Azure region.
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 avoid selecting a model or deployment type unsupported in the intended Azure region.
Question 64
The acceptance criteria for a Foundry-based application require it to invoke a specific deployed model from application code when a catalog model has multiple deployments. Which design should be approved?
- Separate deployment names from model names
- Use the project's scoped endpoint and identity
- Promote infrastructure with declarative configuration
- Use deployment aliases or versioned configuration for controlled rollout
Explanation
Correct Answer
A:For the stated requirement 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 is the mechanism that the system must implement.
Incorrect Answers
B:For a system whose objective is 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 does not directly meet the required outcome to invoke a specific deployed model from application code when a catalog model has multiple deployments.
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 invoke a specific deployed model from application code when a catalog model has multiple deployments.
D: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 invoke a specific deployed model from application code when a catalog model has multiple deployments.
Question 65
After deployment, the team observes: A nightly analytics pipeline pays for interactive latency that none of its downstream consumers need. Which remediation addresses the likely cause?
- Switch the pilot to an eligible Global Standard deployment and monitor metered usage
- Add representative quality and safety evaluations as required checks before production promotion
- Choose a supported regional provisioned deployment and size PTUs using workload measurements
- Submit eligible workloads to a batch deployment and consume completed output asynchronously
Explanation
Correct Answer
D:The fault signal is that a nightly analytics pipeline pays for interactive latency that none of its downstream consumers need; the targeted fix works because batch deployments are designed for large asynchronous jobs and favorable economics; they are not a synchronous low-latency chat endpoint, restoring the ability to summarize a nightly backlog of records without any requirement for immediate responses.
Incorrect Answers
A:If the failing function were to serve an unpredictable low-to-medium traffic pilot without geographic processing restrictions, this change could be relevant because global Standard offers pay-per-token inference with flexible global routing; it does not reserve a fixed amount of dedicated throughput; the actual symptom is that a nightly analytics pipeline pays for interactive latency that none of its downstream consumers need, which calls for a different repair.
B: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 a nightly analytics pipeline pays for interactive latency that none of its downstream consumers need, which calls for a different repair.
C:If the failing function were to support high predictable usage while requiring processing within an Azure geography, this change could be relevant because regional provisioned deployments reserve model throughput within an Azure geography where supported; this couples residency with more predictable capacity; the actual symptom is that a nightly analytics pipeline pays for interactive latency that none of its downstream consumers need, which calls for a different repair.
Question 66
Before approving the proposed action to rotate compromised credentials without app downtime, which measurable acceptance check should be required?
- Compare indexed versions and document counts with the authoritative document repository
- Validate authorized queries succeed and index modification is still denied
- Measure cache-hit ratio, stale-answer rate, and token savings after invalidation tests
- Test service configuration calls and document searches with each role independently
- Confirm no requests use the retired credential and access continues during rotation
Explanation
Correct Answer
E:The accepted change is to rotate compromised credentials without app downtime; this evaluation checks whether it can replace an exposed secret while keeping a production inference endpoint available, since credential rollover with overlapping validity and centralized secret distribution enables staged updates; longer-term managed identity can reduce secret use.
Incorrect Answers
A:This test concerns the ability to detect when newly approved source documents are not available to the grounding pipeline, for which indexer execution status, document counts, and source timestamps show whether retrieval data is current; a good model response cannot compensate for missing content; it does not establish whether the change to rotate compromised credentials without app downtime works for the current release.
B:This test concerns the ability to allow a retrieval application to query only the approved search resource without subscription administration rights, for which Azure RBAC can grant scoped data-plane permissions to the workload principal; broad management roles are unnecessary for query-only service calls; it does not establish whether the change to rotate compromised credentials without app downtime works for the current release.
C:This test concerns the ability to avoid repeating the same approved reference answer for unchanged public content, for which an application cache can reduce repeated inference requests when answers and policy versions are safe to reuse; invalidation and access scope remain essential; it does not establish whether the change to rotate compromised credentials without app downtime works for the current release.
D:This test concerns the ability to let operators change infrastructure while limiting which identities can read sensitive index documents, for which management-plane roles and data-plane roles authorize different operations; control over deployment configuration does not imply every content read permission; it does not establish whether the change to rotate compromised credentials without app downtime works for the current release.
Question 67
A new Foundry-based application must let operators change infrastructure while limiting which identities can read sensitive index documents. Which approach should the team select first?
- Trace each retrieval and generation span
- Restrict agent outbound tool access
- Rotate compromised credentials without app downtime
- Separate data-plane use from resource management
Explanation
Correct Answer
D:For the stated requirement to let operators change infrastructure while limiting which identities can read sensitive index documents, management-plane roles and data-plane roles authorize different operations; control over deployment configuration does not imply every content read permission; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to locate whether a slow RAG response is delayed by search, tool calls, or model generation, distributed traces expose timing and relationships among retrieval, tool and inference spans; aggregate CPU graphs alone cannot isolate these steps; this does not directly meet the required outcome to let operators change infrastructure while limiting which identities can read sensitive index documents.
B:For a system whose objective is to prevent a compromised agent from calling unapproved external endpoints, tool allowlists, network egress policies, and scoped identities limit what an agent can reach; prompt instructions alone are not a network control; this does not directly meet the required outcome to let operators change infrastructure while limiting which identities can read sensitive index documents.
C:For a system whose objective is to replace an exposed secret while keeping a production inference endpoint available, credential rollover with overlapping validity and centralized secret distribution enables staged updates; longer-term managed identity can reduce secret use; this does not directly meet the required outcome to let operators change infrastructure while limiting which identities can read sensitive index documents.
Question 68
An engineer receives this incident report: Model requests return successfully but order lookup tasks fail intermittently at a downstream service. What should be changed?
- Review quotas and reallocate or request capacity for the applicable model and region
- Instrument each tool call with duration, status, and bounded retry decisions
- Align query embeddings with index embeddings and re-embed stored content when migrating models
- Track input and output tokens per step and trim stale context without dropping required facts
Explanation
Correct Answer
B:The fault signal is that model requests return successfully but order lookup tasks fail intermittently at a downstream service; the targeted fix works because tool telemetry records HTTP failures, permission denials and timeouts at action boundaries; model token metrics will not diagnose a downstream API error, restoring the ability to find why agents fail to complete workflows despite healthy model endpoints.
Incorrect Answers
A:If the failing function were to serve a stable higher request rate when the deployment repeatedly reaches its assigned token budget, this change could be relevant because tokens-per-minute and request limits constrain throughput; allocation or capacity decisions must match the deployment's actual measured demand; the actual symptom is that model requests return successfully but order lookup tasks fail intermittently at a downstream service, which calls for a different repair.
C:If the failing function were to catch incompatible document and query vectors when retrieval quality suddenly collapses, this change could be relevant because embedding dimensionality and model compatibility must be maintained between indexed passages and query vectors; vector space changes require a coordinated migration; the actual symptom is that model requests return successfully but order lookup tasks fail intermittently at a downstream service, which calls for a different repair.
D:If the failing function were to identify waste when conversation history and tool outputs grow over successive exchanges, this change could be relevant because per-turn token accounting reveals input-context inflation; controlled summarization or trimming can reduce cost while preserving vital state; the actual symptom is that model requests return successfully but order lookup tasks fail intermittently at a downstream service, which calls for a different repair.
Question 69
Production telemetry reveals this failure: Moderation blocks increase after a client update, but ordinary latency monitoring hides the cause. Which change most directly corrects it?
- Introduce a new credential through a secret store, roll replicas, then revoke the exposed credential
- Add correlated request tracing across retrieval, tools, and model calls
- Run source-grounded evaluations and inspect failing claim-to-passage alignments
- Collect and classify guardrail outcomes with privacy-aware security telemetry
Explanation
Correct Answer
D:The fault signal is that moderation blocks increase after a client update, but ordinary latency monitoring hides the cause; the targeted fix works because guardrail and content-filter telemetry reveals policy-related events; it should be segmented without exposing sensitive payloads unnecessarily, restoring the ability to identify spikes in unsafe or blocked prompts and unsafe generated material.
Incorrect Answers
A:If the failing function were to replace an exposed secret while keeping a production inference endpoint available, this change could be relevant because credential rollover with overlapping validity and centralized secret distribution enables staged updates; longer-term managed identity can reduce secret use; the actual symptom is that moderation blocks increase after a client update, but ordinary latency monitoring hides the cause, which calls for a different repair.
B:If the failing function were to locate whether a slow RAG response is delayed by search, tool calls, or model generation, this change could be relevant because distributed traces expose timing and relationships among retrieval, tool and inference spans; aggregate CPU graphs alone cannot isolate these steps; the actual symptom is that moderation blocks increase after a client update, but ordinary latency monitoring hides the cause, which calls for a different repair.
C:If the failing function were to detect fluent answers that introduce claims unsupported by retrieved passages, this change could be relevant because groundedness evaluations compare generated claims with supplied evidence; fluency or user satisfaction alone cannot establish factual support; the actual symptom is that moderation blocks increase after a client update, but ordinary latency monitoring hides the cause, which calls for a different repair.
Question 70
A development group must find why agents fail to complete workflows despite healthy model endpoints. What is the most appropriate Azure AI implementation?
- Track tool invocation errors and timeouts
- Trace each retrieval and generation span
- Monitor token utilization across agent turns
- Compare behavior across model or prompt versions
Explanation
Correct Answer
A:For the stated requirement to find why agents fail to complete workflows despite healthy model endpoints, tool telemetry records HTTP failures, permission denials and timeouts at action boundaries; model token metrics will not diagnose a downstream API error; this is the mechanism that the system must implement.
Incorrect Answers
B:For a system whose objective is to locate whether a slow RAG response is delayed by search, tool calls, or model generation, distributed traces expose timing and relationships among retrieval, tool and inference spans; aggregate CPU graphs alone cannot isolate these steps; this does not directly meet the required outcome to find why agents fail to complete workflows despite healthy model endpoints.
C:For a system whose objective is to identify waste when conversation history and tool outputs grow over successive exchanges, per-turn token accounting reveals input-context inflation; controlled summarization or trimming can reduce cost while preserving vital state; this does not directly meet the required outcome to find why agents fail to complete workflows despite healthy model endpoints.
D:For a system whose objective is to detect regression when a newly deployed model changes how it answers the same business questions, version-tagged test results separate model changes from request-mix changes and allow a reliable rollback decision; this does not directly meet the required outcome to find why agents fail to complete workflows despite healthy model endpoints.