Microsoft AI-300 Practice Test 38
Question 370
As an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: A registered MLflow model loads on a developer laptop but fails at Azure ML endpoint startup because its conda dependencies and model signature were not packaged with the artifact. The resolution must preserve quality gates, documented artifacts and rollback options. What should the engineer examine first to confirm the specific AI-300 workload issue?
- Inspect MLflow model directory, MLmodel flavors, inferred signature, environment requirements and endpoint container startup traceback.
- Inspect evaluation artifact content, privacy classification, report permissions, explanation samples and downstream dashboard export settings.
- Inspect Python RNG seeding, framework deterministic settings, dataset split creation, logged run parameters and artifact hashes.
- Examine model registry history, artifact retention rules, endpoint references, promotion record and rollback runbook.
Explanation
Correct Answer
A:Inspect MLflow model directory, MLmodel flavors, inferred signature, environment requirements and endpoint container startup traceback. This is the most relevant evidence for mlflow registry artifact cannot load on serving target. MLflow model registration is not proof of portability; required execution environment and signature must accompany the model artifact. The matching controlled remediation is: Register a supported MLflow model with validated flavor, explicit input signature, pinned dependencies and a successful clean-environment load test. This matters here because as an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: A registered….
Incorrect Answers
B:Inspect evaluation artifact content, privacy classification, report permissions, explanation samples and downstream dashboard export settings. This would investigate model evaluation exposes sensitive examples in explainability report rather than the documented mlflow registry artifact cannot load on serving target issue. The relevant investigation is: Inspect MLflow model directory, MLmodel flavors, inferred signature, environment requirements and endpoint container startup traceback. This matters here because as an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: A registered….
C:Inspect Python RNG seeding, framework deterministic settings, dataset split creation, logged run parameters and artifact hashes. This would investigate training script ignores reproducible seed and input contract rather than the documented mlflow registry artifact cannot load on serving target issue. The relevant investigation is: Inspect MLflow model directory, MLmodel flavors, inferred signature, environment requirements and endpoint container startup traceback.
D:Examine model registry history, artifact retention rules, endpoint references, promotion record and rollback runbook. This would investigate incident rollback artifact was deleted instead of archived rather than the documented mlflow registry artifact cannot load on serving target issue. The relevant investigation is: Inspect MLflow model directory, MLmodel flavors, inferred signature, environment requirements and endpoint container startup traceback.
Question 371
During a Microsoft Foundry generative AI deployment review, the following condition occurs: A new troubleshooting prompt sounds polished in sample responses but fails to give actionable next steps because the evaluation measures only fluency and omits task completion and grounding. The deployment must retain governed identities, endpoint contracts and auditable release history. Which production-safe remediation should the engineer apply to this AI workflow?
- Define a versioned structured response instruction and supported JSON schema contract with parser-based validation.
- Permit only the required documented tool and service destinations via supported private or managed-network egress controls.
- Compare candidate prompts using task-specific relevance and groundedness metrics alongside fluency before promotion.
- Evaluate both prompt variants against the same pinned model and representative dataset with controlled assignment and consistent metrics.
Explanation
Correct Answer
C:Compare candidate prompts using task-specific relevance and groundedness metrics alongside fluency before promotion. This targets prompt variant comparison measures fluency not task success. A fluent response can still be wrong or unhelpful; prompt tuning requires evaluation metrics tied to application goals. Verify the failure through: Review prompt variant outputs, evaluator criteria, human task labels, relevance/groundedness scores and business acceptance thresholds. Validate success with: The selected prompt improves successful issue resolution without factual regressions on the held-out support dataset.
Incorrect Answers
A:Define a versioned structured response instruction and supported JSON schema contract with parser-based validation. This could address prompt omits explicit structured answer schema, but it does not solve the specified prompt variant comparison measures fluency not task success failure. The matching remediation is: Compare candidate prompts using task-specific relevance and groundedness metrics alongside fluency before promotion.
B:Permit only the required documented tool and service destinations via supported private or managed-network egress controls. This could address agent tool calls fail behind overbroad outbound block, but it does not solve the specified prompt variant comparison measures fluency not task success failure. The matching remediation is: Compare candidate prompts using task-specific relevance and groundedness metrics alongside fluency before promotion.
D:Evaluate both prompt variants against the same pinned model and representative dataset with controlled assignment and consistent metrics. This could address prompt a/b test uses different model versions in each arm, but it does not solve the specified prompt variant comparison measures fluency not task success failure. The matching remediation is: Compare candidate prompts using task-specific relevance and groundedness metrics alongside fluency before promotion.
Question 372
While preparing a secure Foundry model or agent release, the following condition occurs: A loan-support chatbot produces less useful answers for one customer cohort because all examples in its template come from a narrow set of customer situations. The remediation must meet enterprise access and reliability requirements without a broad bypass. What should the engineer examine first to confirm the specific AI-300 workload issue?
- Examine prompt example coverage, cohort-specific evaluation metrics, response error patterns and real query distribution.
- Review prompt variant outputs, evaluator criteria, human task labels, relevance/groundedness scores and business acceptance thresholds.
- Examine prompt role separation, retrieval snippets, untrusted-content boundary, tool permissions and injection regression tests.
- Review language-specific acceptance tests, deployed model support, retrieval data language coverage, candidate evaluation results and safety requirements.
Explanation
Correct Answer
A:Examine prompt example coverage, cohort-specific evaluation metrics, response error patterns and real query distribution. This is the most relevant evidence for few-shot prompt examples bias outputs toward one subgroup. Prompt variants can embed unrepresentative examples that change behavior; dataset diversity must be checked as carefully as model quality. The matching controlled remediation is: Design and test balanced representative prompt examples under fairness and privacy constraints, retaining a versioned evaluation baseline.
Incorrect Answers
B:Review prompt variant outputs, evaluator criteria, human task labels, relevance/groundedness scores and business acceptance thresholds. This would investigate prompt variant comparison measures fluency not task success rather than the documented few-shot prompt examples bias outputs toward one subgroup issue. The relevant investigation is: Examine prompt example coverage, cohort-specific evaluation metrics, response error patterns and real query distribution.
C:Examine prompt role separation, retrieval snippets, untrusted-content boundary, tool permissions and injection regression tests. This would investigate prompt includes untrusted retrieved instructions as system policy rather than the documented few-shot prompt examples bias outputs toward one subgroup issue. The relevant investigation is: Examine prompt example coverage, cohort-specific evaluation metrics, response error patterns and real query distribution.
D:Review language-specific acceptance tests, deployed model support, retrieval data language coverage, candidate evaluation results and safety requirements. This would investigate multilingual model selection fails target-language quality rather than the documented few-shot prompt examples bias outputs toward one subgroup issue. The relevant investigation is: Examine prompt example coverage, cohort-specific evaluation metrics, response error patterns and real query distribution.
Question 373
During a Microsoft Foundry generative AI deployment review, the following condition occurs: A loan-support chatbot produces less useful answers for one customer cohort because all examples in its template come from a narrow set of customer situations. The deployment must retain governed identities, endpoint contracts and auditable release history. What measured outcome validates the solution while preserving the stated constraints?
- Quality improves on relevant cohorts without exposing personal information or degrading approved common-case answers.
- Every production prompt version resolves to a reviewed commit and a failed change can be rolled back without copying an unknown portal draft.
- The selected prompt improves successful issue resolution without factual regressions on the held-out support dataset.
- Requests from the approved VNet resolve the private address and succeed, while public connectivity remains blocked.
Explanation
Correct Answer
A:Quality improves on relevant cohorts without exposing personal information or degrading approved common-case answers. Prompt variants can embed unrepresentative examples that change behavior; dataset diversity must be checked as carefully as model quality. a direct corrective action is to design and test balanced representative prompt examples under fairness and privacy constraints, retaining a versioned evaluation baseline.
Incorrect Answers
B:Every production prompt version resolves to a reviewed commit and a failed change can be rolled back without copying an unknown portal draft. That addresses a different configuration issue rather than the main requirement. Verify instead: Quality improves on relevant cohorts without exposing personal information or degrading approved common-case answers.
C:The selected prompt improves successful issue resolution without factual regressions on the held-out support dataset. That addresses a different configuration issue rather than the main requirement. Verify instead: Quality improves on relevant cohorts without exposing personal information or degrading approved common-case answers.
D:Requests from the approved VNet resolve the private address and succeed, while public connectivity remains blocked. That addresses a different configuration issue rather than the main requirement. Verify instead: Quality improves on relevant cohorts without exposing personal information or degrading approved common-case answers.
Question 374
While preparing a secure Foundry model or agent release, the following condition occurs: An agent receives a newly released model under full production traffic before anyone verifies tool calling and safety regressions. The remediation must meet enterprise access and reliability requirements without a broad bypass. Which investigation best verifies the cause before changing production configuration?
- Inspect Foundry project and deployment inventory, pipeline environment variables, identities, API endpoints and change audit logs.
- Examine billed PTU hours, token utilization, workload predictability, available pay-as-you-go endpoints and expected launch load.
- Compare real tokenized input sizes, selected model version limits, deployment configuration and representative failure logs.
- Inspect deployment routing weights, evaluated model versions, incident timing, canary quality dashboards and rollback path.
Explanation
Correct Answer
D:Inspect deployment routing weights, evaluated model versions, incident timing, canary quality dashboards and rollback path. This is the most relevant evidence for new foundation model replaces production without canary. Gradual rollout with a known-good model mitigates production blast radius when foundation model behavior changes. The matching controlled remediation is: Route a controlled percentage of requests to the candidate and promote only after automated quality, safety and latency gates pass.
Incorrect Answers
A:Inspect Foundry project and deployment inventory, pipeline environment variables, identities, API endpoints and change audit logs. This would investigate foundry environment mixes development and production agent resources rather than the documented new foundation model replaces production without canary issue. The relevant investigation is: Inspect deployment routing weights, evaluated model versions, incident timing, canary quality dashboards and rollback path.
B:Examine billed PTU hours, token utilization, workload predictability, available pay-as-you-go endpoints and expected launch load. This would investigate provisioned deployment pays for idle unused capacity rather than the documented new foundation model replaces production without canary issue. The relevant investigation is: Inspect deployment routing weights, evaluated model versions, incident timing, canary quality dashboards and rollback path.
C:Compare real tokenized input sizes, selected model version limits, deployment configuration and representative failure logs. This would investigate foundation model selected without required context window rather than the documented new foundation model replaces production without canary issue. The relevant investigation is: Inspect deployment routing weights, evaluated model versions, incident timing, canary quality dashboards and rollback path.
Question 375
During a regulated Azure ML workspace infrastructure review, the following condition occurs: An engineer runs an Azure CLI setup script that creates the model workspace in the CLI's current default subscription rather than the approved production subscription. The team must preserve approved access, model provenance and private network constraints. Which production-safe remediation should the engineer apply to this AI workflow?
- Configure GitHub OIDC federated identity with a scoped Azure service principal and replace the long-lived secret in the deployment workflow.
- Link the correct private DNS zone and approved storage private endpoint to the managed network path before enforcing storage isolation.
- Require explicit subscription and resource group targeting and preflight assertions before executing repeatable workspace deployment commands.
- Replace embedded workspace-specific paths with parameterized, typed data inputs and bind them explicitly in each consuming pipeline.
Explanation
Correct Answer
C:Require explicit subscription and resource group targeting and preflight assertions before executing repeatable workspace deployment commands. This targets azure cli deployment silently targets sandbox defaults. Azure CLI context is stateful; an otherwise valid infrastructure command can target the wrong subscription when scope is not explicitly established. Verify the failure through: Inspect az account subscription context, deployment scope flags, tenant identity, Bicep parameter file and cloud audit evidence. Validate success with: Production resources appear only in the authorized scope and rerunning the script does not create duplicate sandbox resources. This matters here because an engineer runs an Azure CLI setup script that creates the model workspace in the CLI's current default….
Incorrect Answers
A:Configure GitHub OIDC federated identity with a scoped Azure service principal and replace the long-lived secret in the deployment workflow. This could address github action uses persistent azure client secret, but it does not solve the specified azure cli deployment silently targets sandbox defaults failure. The matching remediation is: Require explicit subscription and resource group targeting and preflight assertions before executing repeatable workspace deployment commands.
B:Link the correct private DNS zone and approved storage private endpoint to the managed network path before enforcing storage isolation. This could address aml private workspace storage access fails from compute, but it does not solve the specified azure cli deployment silently targets sandbox defaults failure. The matching remediation is: Require explicit subscription and resource group targeting and preflight assertions before executing repeatable workspace deployment commands.
D:Replace embedded workspace-specific paths with parameterized, typed data inputs and bind them explicitly in each consuming pipeline. This could address reusable component hardcodes workspace-specific asset path, but it does not solve the specified azure cli deployment silently targets sandbox defaults failure. The matching remediation is: Require explicit subscription and resource group targeting and preflight assertions before executing repeatable workspace deployment commands.
Question 376
As an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: A deployed churn model receives increasing nulls in a new CRM field but its data drift monitor tracks only average values and ignores missingness and schema validity. The resolution must preserve quality gates, documented artifacts and rollback options. Which observable postchange result proves the intended remediation succeeded?
- A poor candidate is retained for review without reaching production, and qualifying candidates progress through controlled canary rollout.
- Reports distinguish provisional drift/latency signals from later evidence-based accuracy and recall.
- An injected null-rate increase triggers a targeted alert while valid fluctuations remain within control limits.
- A controlled API load test meets latency and correctness targets without requiring clients to poll batch jobs.
Explanation
Correct Answer
C:An injected null-rate increase triggers a targeted alert while valid fluctuations remain within control limits. Data quality and data distribution drift are related but distinct signals; averages alone can conceal feature loss. a direct corrective action is to add suitable data quality and schema drift signals with thresholds for missingness alongside distribution drift.
Incorrect Answers
A:A poor candidate is retained for review without reaching production, and qualifying candidates progress through controlled canary rollout. That addresses a different configuration issue rather than the main requirement. Verify instead: An injected null-rate increase triggers a targeted alert while valid fluctuations remain within control limits.
B:Reports distinguish provisional drift/latency signals from later evidence-based accuracy and recall. That addresses a different configuration issue rather than the main requirement. Verify instead: An injected null-rate increase triggers a targeted alert while valid fluctuations remain within control limits.
D:A controlled API load test meets latency and correctness targets without requiring clients to poll batch jobs. That addresses a different configuration issue rather than the main requirement. Verify instead: An injected null-rate increase triggers a targeted alert while valid fluctuations remain within control limits.
Question 377
During a production Azure Machine Learning model lifecycle incident, the following condition occurs: A nightly 15-million-row customer scoring task is sent as individual requests to a production real-time endpoint, degrading its daytime customer traffic. The operator must safeguard training reproducibility and serving continuity. Which TWO complementary choices identify the issue and correct it? Select TWO.
- Review deployment container logs, init() traceback, model mount path, environment image and readiness/liveness probe results.
- Bind training input directly to the feature component's typed output to establish the correct pipeline dependency.
- Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability.
- Align the published endpoint contract and client payload mapping, then deploy the reviewed scoring code and test representative requests.
- Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
Explanation
Correct Answers
C:Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability. This is the most relevant evidence for overnight bulk scoring saturates interactive online endpoint. Batch scoring and synchronous online inference serve different workload shapes and should not contend for capacity unnecessarily. The matching controlled remediation is: Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
E:Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls. This targets overnight bulk scoring saturates interactive online endpoint. Batch scoring and synchronous online inference serve different workload shapes and should not contend for capacity unnecessarily. Verify the failure through: Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability. Validate success with: Overnight processing completes within its processing window and interactive endpoint latency remains within SLA.
Incorrect Answers
A:Review deployment container logs, init() traceback, model mount path, environment image and readiness/liveness probe results. This is associated with endpoint deployment health probe fails before model initialization, not the essential evidence-and-fix pair for overnight bulk scoring saturates interactive online endpoint. First examine: Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability. Then correct: Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
B:Bind training input directly to the feature component's typed output to establish the correct pipeline dependency. This is associated with pipeline graph omits feature-build dependency, not the essential evidence-and-fix pair for overnight bulk scoring saturates interactive online endpoint. First examine: Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability. Then correct: Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
D:Align the published endpoint contract and client payload mapping, then deploy the reviewed scoring code and test representative requests. This is associated with endpoint scoring script expects wrong json schema, not the essential evidence-and-fix pair for overnight bulk scoring saturates interactive online endpoint. First examine: Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability. Then correct: Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
Question 378
While preparing a secure Foundry model or agent release, the following condition occurs: A custom transformer model is deployed on managed compute but inference containers crash on startup because the image's tokenizer library is incompatible with the model artifact. The remediation must meet enterprise access and reliability requirements without a broad bypass. Which investigation best verifies the cause before changing production configuration?
- Examine model package runtime requirements, container startup traceback, tokenizer/config files and managed compute environment image digest.
- Compare model deployment version pinning, response schemas, regression dataset, rollout timeline and application parser errors.
- Inspect workflow identity claims, login failure logs, secret references, federated identity options and target resource permissions.
- Examine billed PTU hours, token utilization, workload predictability, available pay-as-you-go endpoints and expected launch load.
Explanation
Correct Answer
A:Examine model package runtime requirements, container startup traceback, tokenizer/config files and managed compute environment image digest. This is the most relevant evidence for managed foundation model deployment lacks compatible runtime image. Managed inference still requires the model weights and runtime dependencies to be compatible; compute provisioning cannot fix a broken model-serving image. The matching controlled remediation is: Build and test a compatible version-pinned serving environment and promote the corrected managed model deployment. This matters here because while preparing a secure Foundry model or agent release, the following condition occurs: A custom transformer model….
Incorrect Answers
B:Compare model deployment version pinning, response schemas, regression dataset, rollout timeline and application parser errors. This would investigate foundry deployment alias changes model behavior unexpectedly rather than the documented managed foundation model deployment lacks compatible runtime image issue. The relevant investigation is: Examine model package runtime requirements, container startup traceback, tokenizer/config files and managed compute environment image digest.
C:Inspect workflow identity claims, login failure logs, secret references, federated identity options and target resource permissions. This would investigate foundry service principal credential rotated without updating pipeline rather than the documented managed foundation model deployment lacks compatible runtime image issue. The relevant investigation is: Examine model package runtime requirements, container startup traceback, tokenizer/config files and managed compute environment image digest.
D:Examine billed PTU hours, token utilization, workload predictability, available pay-as-you-go endpoints and expected launch load. This would investigate provisioned deployment pays for idle unused capacity rather than the documented managed foundation model deployment lacks compatible runtime image issue. The relevant investigation is: Examine model package runtime requirements, container startup traceback, tokenizer/config files and managed compute environment image digest.
Question 379
During a production Azure Machine Learning model lifecycle incident, the following condition occurs: An AutoML classification search reports near-perfect accuracy because the supplied training dataset includes a field recorded only after the real customer outcome is known. The operator must safeguard training reproducibility and serving continuity. Which observable postchange result proves the intended remediation succeeded?
- Holdout performance is credible against production-like inputs and candidate selection no longer uses post-outcome fields.
- The promotion report presents comparable metrics with documented confidence and class distribution before model approval.
- Local notebook and managed job both import the same pinned dependency and generate comparable model outputs.
- A poor candidate is retained for review without reaching production, and qualifying candidates progress through controlled canary rollout.
Explanation
Correct Answer
A:Holdout performance is credible against production-like inputs and candidate selection no longer uses post-outcome fields. Automated model search cannot correct information leakage caused by improperly constructed input data and validation boundaries. a direct corrective action is to remove target-leakage columns and rerun the AutoML search with an appropriate time-aware validation split and realistic available features.
Incorrect Answers
B:The promotion report presents comparable metrics with documented confidence and class distribution before model approval. That addresses a different configuration issue rather than the main requirement. Verify instead: Holdout performance is credible against production-like inputs and candidate selection no longer uses post-outcome fields.
C:Local notebook and managed job both import the same pinned dependency and generate comparable model outputs. That addresses a different configuration issue rather than the main requirement. Verify instead: Holdout performance is credible against production-like inputs and candidate selection no longer uses post-outcome fields.
D:A poor candidate is retained for review without reaching production, and qualifying candidates progress through controlled canary rollout. That addresses a different configuration issue rather than the main requirement. Verify instead: Holdout performance is credible against production-like inputs and candidate selection no longer uses post-outcome fields.
Question 380
During a production Azure Machine Learning model lifecycle incident, the following condition occurs: A model is rolled back to old weights but its scoring container still uses the new incompatible environment, so production errors continue after rollback. The operator must safeguard training reproducibility and serving continuity. Select TWO: the strongest initial evidence and the aligned controlled correction.
- Compare candidate and previous deployment model version, environment digest, scoring code, endpoint configuration and historical release manifest.
- Deploy the approved model through a suitable managed online endpoint and provision capacity for synchronous requests.
- Capture a governed immutable dataset or versioned data asset and reference that exact version in notebook and scheduled jobs.
- Inspect online endpoint traffic weights, deployment versions, request tracing, rollout approval and available prior stable deployment.
- Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights.
Explanation
Correct Answers
A:Compare candidate and previous deployment model version, environment digest, scoring code, endpoint configuration and historical release manifest. This is the most relevant evidence for rollback cannot recover a previous environment version. A deployment's behavior depends on the whole inference package; changing model weights alone cannot undo incompatible runtime changes. The matching controlled remediation is: Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights.
E:Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights. This targets rollback cannot recover a previous environment version. A deployment's behavior depends on the whole inference package; changing model weights alone cannot undo incompatible runtime changes. Verify the failure through: Compare candidate and previous deployment model version, environment digest, scoring code, endpoint configuration and historical release manifest. Validate success with: A representative rollback test restores known-good predictions and endpoint health within the recovery target.
Incorrect Answers
B:Deploy the approved model through a suitable managed online endpoint and provision capacity for synchronous requests. This is associated with online scoring requirement deployed as batch endpoint, not the essential evidence-and-fix pair for rollback cannot recover a previous environment version. First examine: Compare candidate and previous deployment model version, environment digest, scoring code, endpoint configuration and historical release manifest. Then correct: Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights.
C:Capture a governed immutable dataset or versioned data asset and reference that exact version in notebook and scheduled jobs. This is associated with exploration notebook reads production data without reproducible snapshot, not the essential evidence-and-fix pair for rollback cannot recover a previous environment version. First examine: Compare candidate and previous deployment model version, environment digest, scoring code, endpoint configuration and historical release manifest. Then correct: Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights.
D:Inspect online endpoint traffic weights, deployment versions, request tracing, rollout approval and available prior stable deployment. This is associated with canary rollout sends all traffic to unvalidated candidate, not the essential evidence-and-fix pair for rollback cannot recover a previous environment version. First examine: Compare candidate and previous deployment model version, environment digest, scoring code, endpoint configuration and historical release manifest. Then correct: Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights.