McAfee-Secured Website

Exam Code: AI-300

Exam Name: Operationalizing Machine Learning and Generative AI Solutions

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: Machine Learning Operations Engineer Associate

100% Microsoft AI-300 Guaranteed Success With Testking Training Materials

Get Real AI-300 Exam Questions & Expert Verified Answers as Experienced on the Real Exam!

213 Questions & Answers with Testing Engine

"Operationalizing Machine Learning and Generative AI Solutions Exam", also known as AI-300 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date AI-300 Exam Engine. Your AI-300 training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Microsoft AI-300 Practice Test 4

Question 30

A job logs under an employee's identity on a local compute instance but fails on a managed cluster when reading storage. Which setting should be inspected?

  1. Request the correct VM-family core quota or select another permitted SKU.
  2. Use an autoscaling compute cluster with minimum node count set to zero.
  3. Assign and grant data access to the appropriate managed identity for the compute execution.
  4. Choose a supported GPU compute target with the necessary GPU quota and runtime.

Explanation

Correct Answer

C:

Job identity can differ from interactive user identity; runtime access must be granted explicitly to the executing context.

Incorrect Answers

A:

The alternative, Request the correct VM-family core quota or select another permitted SKU, belongs to a different technical condition: A newly created high-memory training cluster stays at zero capacity after all jobs enter the queue. The quota page reports zero remaining cores in that VM family. There the source-specific mechanism or proof is: Cluster scaling is bounded by regional and VM-family quota; queued work cannot create unavailable cores automatically. The condition asked about here is: A job logs under an employee's identity on a local compute instance but fails on a managed cluster when reading storage. A suitable correction is Assign and grant data access to the appropriate managed identity for the compute execution; the targeted justification or result is: Job identity can differ from interactive user identity; runtime access must be granted explicitly to the executing context.

B:

Use an autoscaling compute cluster with minimum node count set to zero. That operation would be relevant if: A team submits many short training jobs and wants nodes to scale down to zero when no jobs remain. Its corresponding reason or verification is: Managed compute clusters can provision capacity for jobs and scale back when idle, reducing unnecessary compute charges. However the problem in this question is: A job logs under an employee's identity on a local compute instance but fails on a managed cluster when reading storage. The requirement instead calls for Assign and grant data access to the appropriate managed identity for the compute execution; the question-specific check is: Job identity can differ from interactive user identity; runtime access must be granted explicitly to the executing context.

D:

Choose a supported GPU compute target with the necessary GPU quota and runtime. A separate source case supports that approach when: A deep-learning job cannot start on the selected CPU-only cluster because CUDA acceleration is required. The supporting rationale or observable result is: GPU training needs compatible hardware and drivers; changing the model display name cannot provide CUDA capacity. This question describes a different issue: A job logs under an employee's identity on a local compute instance but fails on a managed cluster when reading storage. Address it through Assign and grant data access to the appropriate managed identity for the compute execution, with the required justification or evidence: Job identity can differ from interactive user identity; runtime access must be granted explicitly to the executing context.

Question 31

A team wants predictable concurrency for 35 simultaneous jobs but its compute cluster maximum nodes is fixed at one. Which property should be reconsidered?

  1. Configure idle shutdown or automatic start/stop schedules for the compute instance.
  2. Increase the cluster's maximum node count within quota and workload capacity requirements.
  3. Point the job to an existing registered compute target or provision its required cluster.
  4. Assign and grant data access to the appropriate managed identity for the compute execution.

Explanation

Correct Answer

B:

Autoscaling capacity limits determine how much work can execute concurrently; a larger job queue alone cannot increase throughput.

Incorrect Answers

A:

Configure idle shutdown or automatic start/stop schedules for the compute instance. It addresses another case: A compute instance remains billed overnight even when no notebook kernel is active. For that case, the distinguishing justification or acceptance check is: Instance billing continues while running; an idle-stop schedule reduces running hours without deleting code and notebooks. The present case instead states: A team wants predictable concurrency for 35 simultaneous jobs but its compute cluster maximum nodes is fixed at one. It calls for Increase the cluster's maximum node count within quota and workload capacity requirements. Showing success for the other case would not establish: Autoscaling capacity limits determine how much work can execute concurrently; a larger job queue alone cannot increase throughput.

C:

The alternative, Point the job to an existing registered compute target or provision its required cluster, belongs to a different technical condition: A cluster was deleted after a completed experiment; a new run still references its name and fails before training starts. There the source-specific mechanism or proof is: Compute target names must resolve to available resources; a preserved experiment record cannot execute on deleted infrastructure. The condition asked about here is: A team wants predictable concurrency for 35 simultaneous jobs but its compute cluster maximum nodes is fixed at one. A suitable correction is Increase the cluster's maximum node count within quota and workload capacity requirements; the targeted justification or result is: Autoscaling capacity limits determine how much work can execute concurrently; a larger job queue alone cannot increase throughput.

D:

Assign and grant data access to the appropriate managed identity for the compute execution. That operation would be relevant if: A job logs under an employee's identity on a local compute instance but fails on a managed cluster when reading storage. Its corresponding reason or verification is: Job identity can differ from interactive user identity; runtime access must be granted explicitly to the executing context. However the problem in this question is: A team wants predictable concurrency for 35 simultaneous jobs but its compute cluster maximum nodes is fixed at one. The requirement instead calls for Increase the cluster's maximum node count within quota and workload capacity requirements; the question-specific check is: Autoscaling capacity limits determine how much work can execute concurrently; a larger job queue alone cannot increase throughput.

Question 32

A large training run repeatedly exhausts memory but CPU utilization remains low on current nodes. What compute decision is most relevant?

  1. Select a compute SKU with sufficient memory per node for the job's actual footprint.
  2. Provision a user-dedicated Azure Machine Learning compute instance for interactive development.
  3. Increase the cluster's maximum node count within quota and workload capacity requirements.
  4. Assign and grant data access to the appropriate managed identity for the compute execution.

Explanation

Correct Answer

A:

Memory pressure depends on VM characteristics and job data; increasing CPU utilization settings does not add physical RAM.

Incorrect Answers

B:

Provision a user-dedicated Azure Machine Learning compute instance for interactive development. This is useful for the distinct issue: A data scientist wants a persistent managed notebook development machine for interactive experiments. The relevant technical rationale or evidence for that issue is: Compute instances provide managed interactive workstations; shared autoscaling job clusters are designed for submitted execution workloads. Here the reported issue is: A large training run repeatedly exhausts memory but CPU utilization remains low on current nodes. The relevant response is: Select a compute SKU with sufficient memory per node for the job's actual footprint. The first result cannot verify this different correction: Memory pressure depends on VM characteristics and job data; increasing CPU utilization settings does not add physical RAM.

C:

Increase the cluster's maximum node count within quota and workload capacity requirements. It addresses another case: A team wants predictable concurrency for 35 simultaneous jobs but its compute cluster maximum nodes is fixed at one. For that case, the distinguishing justification or acceptance check is: Autoscaling capacity limits determine how much work can execute concurrently; a larger job queue alone cannot increase throughput. The present case instead states: A large training run repeatedly exhausts memory but CPU utilization remains low on current nodes. It calls for Select a compute SKU with sufficient memory per node for the job's actual footprint. Showing success for the other case would not establish: Memory pressure depends on VM characteristics and job data; increasing CPU utilization settings does not add physical RAM.

D:

The alternative, Assign and grant data access to the appropriate managed identity for the compute execution, belongs to a different technical condition: A job logs under an employee's identity on a local compute instance but fails on a managed cluster when reading storage. There the source-specific mechanism or proof is: Job identity can differ from interactive user identity; runtime access must be granted explicitly to the executing context. The condition asked about here is: A large training run repeatedly exhausts memory but CPU utilization remains low on current nodes. A suitable correction is Select a compute SKU with sufficient memory per node for the job's actual footprint; the targeted justification or result is: Memory pressure depends on VM characteristics and job data; increasing CPU utilization settings does not add physical RAM.

Question 33

A cluster was deleted after a completed experiment; a new run still references its name and fails before training starts. How should the pipeline be fixed?

  1. Configure idle shutdown or automatic start/stop schedules for the compute instance.
  2. Select a compute SKU with sufficient memory per node for the job's actual footprint.
  3. Increase the cluster's maximum node count within quota and workload capacity requirements.
  4. Point the job to an existing registered compute target or provision its required cluster.

Explanation

Correct Answer

D:

Compute target names must resolve to available resources; a preserved experiment record cannot execute on deleted infrastructure.

Incorrect Answers

A:

Configure idle shutdown or automatic start/stop schedules for the compute instance. A separate source case supports that approach when: A compute instance remains billed overnight even when no notebook kernel is active. The supporting rationale or observable result is: Instance billing continues while running; an idle-stop schedule reduces running hours without deleting code and notebooks. This question describes a different issue: A cluster was deleted after a completed experiment; a new run still references its name and fails before training starts. Address it through Point the job to an existing registered compute target or provision its required cluster, with the required justification or evidence: Compute target names must resolve to available resources; a preserved experiment record cannot execute on deleted infrastructure.

B:

Select a compute SKU with sufficient memory per node for the job's actual footprint. This is useful for the distinct issue: A large training run repeatedly exhausts memory but CPU utilization remains low on current nodes. The relevant technical rationale or evidence for that issue is: Memory pressure depends on VM characteristics and job data; increasing CPU utilization settings does not add physical RAM. Here the reported issue is: A cluster was deleted after a completed experiment; a new run still references its name and fails before training starts. The relevant response is: Point the job to an existing registered compute target or provision its required cluster. The first result cannot verify this different correction: Compute target names must resolve to available resources; a preserved experiment record cannot execute on deleted infrastructure.

C:

Increase the cluster's maximum node count within quota and workload capacity requirements. It addresses another case: A team wants predictable concurrency for 35 simultaneous jobs but its compute cluster maximum nodes is fixed at one. For that case, the distinguishing justification or acceptance check is: Autoscaling capacity limits determine how much work can execute concurrently; a larger job queue alone cannot increase throughput. The present case instead states: A cluster was deleted after a completed experiment; a new run still references its name and fails before training starts. It calls for Point the job to an existing registered compute target or provision its required cluster. Showing success for the other case would not establish: Compute target names must resolve to available resources; a preserved experiment record cannot execute on deleted infrastructure.

Question 34

A junior operator needs to submit ML training jobs but must not change workspace networking or grant user roles. What authorization approach fits?

  1. Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner.
  2. Replace personal-user dependencies with an approved nonhuman workload identity.
  3. Use a dedicated federated or managed workload identity with minimal Azure RBAC.
  4. Assign the appropriate Key Vault data-plane permissions at the secret resource scope.

Explanation

Correct Answer

A:

Least-privilege workspace RBAC permits job operations without broad control over security or billing resources.

Incorrect Answers

B:

Replace personal-user dependencies with an approved nonhuman workload identity. That operation would be relevant if: An access policy depends on an employee's personal account, and the employee leaves the company. Its corresponding reason or verification is: Long-running automation should not rely on a human account whose employment and conditional access status can change. However the problem in this question is: A junior operator needs to submit ML training jobs but must not change workspace networking or grant user roles. The requirement instead calls for Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner; the question-specific check is: Least-privilege workspace RBAC permits job operations without broad control over security or billing resources.

C:

Use a dedicated federated or managed workload identity with minimal Azure RBAC. A separate source case supports that approach when: A GitHub deployment workflow needs access to Azure ML without storing an administrator password. The supporting rationale or observable result is: Workload identity avoids persistent administrator secrets and can be constrained to the specific resources it deploys. This question describes a different issue: A junior operator needs to submit ML training jobs but must not change workspace networking or grant user roles. Address it through Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner, with the required justification or evidence: Least-privilege workspace RBAC permits job operations without broad control over security or billing resources.

D:

Assign the appropriate Key Vault data-plane permissions at the secret resource scope. This is useful for the distinct issue: A new team member receives the expected workspace role yet still cannot access a separately managed Azure Key Vault. The relevant technical rationale or evidence for that issue is: Azure ML role assignments do not automatically authorize secret retrieval in linked services with separate access policies. Here the reported issue is: A junior operator needs to submit ML training jobs but must not change workspace networking or grant user roles. The relevant response is: Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner. The first result cannot verify this different correction: Least-privilege workspace RBAC permits job operations without broad control over security or billing resources.

Question 35

A GitHub deployment workflow needs access to Azure ML without storing an administrator password. What identity choice is preferred?

  1. Give the endpoint deployment's runtime identity the necessary feature-store data permissions.
  2. Remove embedded keys and configure data-plane access for the executing managed identity.
  3. Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner.
  4. Use a dedicated federated or managed workload identity with minimal Azure RBAC.

Explanation

Correct Answer

D:

Workload identity avoids persistent administrator secrets and can be constrained to the specific resources it deploys.

Incorrect Answers

A:

The alternative, Give the endpoint deployment's runtime identity the necessary feature-store data permissions, belongs to a different technical condition: A production endpoint can fetch its model but cannot read a secured feature store at inference time. There the source-specific mechanism or proof is: Production inference uses its execution identity; author privileges do not automatically travel to the deployed container. The condition asked about here is: A GitHub deployment workflow needs access to Azure ML without storing an administrator password. A suitable correction is Use a dedicated federated or managed workload identity with minimal Azure RBAC; the targeted justification or result is: Workload identity avoids persistent administrator secrets and can be constrained to the specific resources it deploys.

B:

Remove embedded keys and configure data-plane access for the executing managed identity. That operation would be relevant if: A revoked storage account key remains in an old training script, although the operator can read storage through their own account. Its corresponding reason or verification is: Interactive user rights cannot rescue a job using revoked credentials; explicit identity-based access is auditable and renewable. However the problem in this question is: A GitHub deployment workflow needs access to Azure ML without storing an administrator password. The requirement instead calls for Use a dedicated federated or managed workload identity with minimal Azure RBAC; the question-specific check is: Workload identity avoids persistent administrator secrets and can be constrained to the specific resources it deploys.

C:

Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner. A separate source case supports that approach when: A junior operator needs to submit ML training jobs but must not change workspace networking or grant user roles. The supporting rationale or observable result is: Least-privilege workspace RBAC permits job operations without broad control over security or billing resources. This question describes a different issue: A GitHub deployment workflow needs access to Azure ML without storing an administrator password. Address it through Use a dedicated federated or managed workload identity with minimal Azure RBAC, with the required justification or evidence: Workload identity avoids persistent administrator secrets and can be constrained to the specific resources it deploys.

Question 36

A production endpoint can fetch its model but cannot read a secured feature store at inference time. Where should authorization be granted?

  1. Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access.
  2. Use a dedicated federated or managed workload identity with minimal Azure RBAC.
  3. Give the endpoint deployment's runtime identity the necessary feature-store data permissions.
  4. Check the user's effective Azure RBAC authorization independently from network connectivity.

Explanation

Correct Answer

C:

Production inference uses its execution identity; author privileges do not automatically travel to the deployed container.

Incorrect Answers

A:

Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access. It addresses another case: A contractor needs read-only access to experiment metrics but no ability to download sensitive datasets. For that case, the distinguishing justification or acceptance check is: Workspace metadata permission does not automatically define every storage grant; separate resource policies protect restricted data. The present case instead states: A production endpoint can fetch its model but cannot read a secured feature store at inference time. It calls for Give the endpoint deployment's runtime identity the necessary feature-store data permissions. Showing success for the other case would not establish: Production inference uses its execution identity; author privileges do not automatically travel to the deployed container.

B:

The alternative, Use a dedicated federated or managed workload identity with minimal Azure RBAC, belongs to a different technical condition: A GitHub deployment workflow needs access to Azure ML without storing an administrator password. There the source-specific mechanism or proof is: Workload identity avoids persistent administrator secrets and can be constrained to the specific resources it deploys. The condition asked about here is: A production endpoint can fetch its model but cannot read a secured feature store at inference time. A suitable correction is Give the endpoint deployment's runtime identity the necessary feature-store data permissions; the targeted justification or result is: Production inference uses its execution identity; author privileges do not automatically travel to the deployed container.

D:

Check the user's effective Azure RBAC authorization independently from network connectivity. That operation would be relevant if: An ML workspace is accessible through a private endpoint, but a notebook user's access is denied despite the correct network path. Its corresponding reason or verification is: Private networking controls reachability, not permission to perform workspace operations; both controls must allow the request. However the problem in this question is: A production endpoint can fetch its model but cannot read a secured feature store at inference time. The requirement instead calls for Give the endpoint deployment's runtime identity the necessary feature-store data permissions; the question-specific check is: Production inference uses its execution identity; author privileges do not automatically travel to the deployed container.

Question 37

An engineer is able to open the workspace portal but receives authorization failures when trying to edit compute. What should be checked first?

  1. Remove embedded keys and configure data-plane access for the executing managed identity.
  2. Examine the exact role assignment and scope for the requested compute operation.
  3. Replace personal-user dependencies with an approved nonhuman workload identity.
  4. Assign the appropriate Key Vault data-plane permissions at the secret resource scope.

Explanation

Correct Answer

B:

Portal visibility and write permissions are distinct capabilities; the user needs the specific role action at the effective resource scope.

Incorrect Answers

A:

Remove embedded keys and configure data-plane access for the executing managed identity. This is useful for the distinct issue: A revoked storage account key remains in an old training script, although the operator can read storage through their own account. The relevant technical rationale or evidence for that issue is: Interactive user rights cannot rescue a job using revoked credentials; explicit identity-based access is auditable and renewable. Here the reported issue is: An engineer is able to open the workspace portal but receives authorization failures when trying to edit compute. The relevant response is: Examine the exact role assignment and scope for the requested compute operation. The first result cannot verify this different correction: Portal visibility and write permissions are distinct capabilities; the user needs the specific role action at the effective resource scope.

C:

Replace personal-user dependencies with an approved nonhuman workload identity. It addresses another case: An access policy depends on an employee's personal account, and the employee leaves the company. For that case, the distinguishing justification or acceptance check is: Long-running automation should not rely on a human account whose employment and conditional access status can change. The present case instead states: An engineer is able to open the workspace portal but receives authorization failures when trying to edit compute. It calls for Examine the exact role assignment and scope for the requested compute operation. Showing success for the other case would not establish: Portal visibility and write permissions are distinct capabilities; the user needs the specific role action at the effective resource scope.

D:

The alternative, Assign the appropriate Key Vault data-plane permissions at the secret resource scope, belongs to a different technical condition: A new team member receives the expected workspace role yet still cannot access a separately managed Azure Key Vault. There the source-specific mechanism or proof is: Azure ML role assignments do not automatically authorize secret retrieval in linked services with separate access policies. The condition asked about here is: An engineer is able to open the workspace portal but receives authorization failures when trying to edit compute. A suitable correction is Examine the exact role assignment and scope for the requested compute operation; the targeted justification or result is: Portal visibility and write permissions are distinct capabilities; the user needs the specific role action at the effective resource scope.

Question 38

A security review finds a service principal with Contributor permission across the subscription, although it deploys models to only one workspace. What should be changed?

  1. Restrict the service principal to the minimal workspace or resource-group role scope.
  2. Assign the appropriate Key Vault data-plane permissions at the secret resource scope.
  3. Replace personal-user dependencies with an approved nonhuman workload identity.
  4. Remove embedded keys and configure data-plane access for the executing managed identity.

Explanation

Correct Answer

A:

Reducing privilege constrains a compromised pipeline identity while retaining only the operations its job requires.

Incorrect Answers

B:

Assign the appropriate Key Vault data-plane permissions at the secret resource scope. A separate source case supports that approach when: A new team member receives the expected workspace role yet still cannot access a separately managed Azure Key Vault. The supporting rationale or observable result is: Azure ML role assignments do not automatically authorize secret retrieval in linked services with separate access policies. This question describes a different issue: A security review finds a service principal with Contributor permission across the subscription, although it deploys models to only one workspace. Address it through Restrict the service principal to the minimal workspace or resource-group role scope, with the required justification or evidence: Reducing privilege constrains a compromised pipeline identity while retaining only the operations its job requires.

C:

Replace personal-user dependencies with an approved nonhuman workload identity. This is useful for the distinct issue: An access policy depends on an employee's personal account, and the employee leaves the company. The relevant technical rationale or evidence for that issue is: Long-running automation should not rely on a human account whose employment and conditional access status can change. Here the reported issue is: A security review finds a service principal with Contributor permission across the subscription, although it deploys models to only one workspace. The relevant response is: Restrict the service principal to the minimal workspace or resource-group role scope. The first result cannot verify this different correction: Reducing privilege constrains a compromised pipeline identity while retaining only the operations its job requires.

D:

Remove embedded keys and configure data-plane access for the executing managed identity. It addresses another case: A revoked storage account key remains in an old training script, although the operator can read storage through their own account. For that case, the distinguishing justification or acceptance check is: Interactive user rights cannot rescue a job using revoked credentials; explicit identity-based access is auditable and renewable. The present case instead states: A security review finds a service principal with Contributor permission across the subscription, although it deploys models to only one workspace. It calls for Restrict the service principal to the minimal workspace or resource-group role scope. Showing success for the other case would not establish: Reducing privilege constrains a compromised pipeline identity while retaining only the operations its job requires.

Question 39

An access policy depends on an employee's personal account, and the employee leaves the company. Which design prevents unattended ML pipelines from losing access?

  1. Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access.
  2. Increase the departed employee account's Azure RBAC scope while keeping that account as the pipeline identity.
  3. Store a long-lived refresh token for the departed personal account in the scheduled pipeline configuration.
  4. Replace personal-user dependencies with an approved nonhuman workload identity.

Explanation

Correct Answer

D:

Long-running automation should not rely on a human account whose employment and conditional access status can change.

Incorrect Answers

A:

Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access. That operation would be relevant if: A contractor needs read-only access to experiment metrics but no ability to download sensitive datasets. Its corresponding reason or verification is: Workspace metadata permission does not automatically define every storage grant; separate resource policies protect restricted data. However the problem in this question is: An access policy depends on an employee's personal account, and the employee leaves the company. The requirement instead calls for Replace personal-user dependencies with an approved nonhuman workload identity; the question-specific check is: Long-running automation should not rely on a human account whose employment and conditional access status can change.

B:

Additional role permissions cannot prevent automation failure when its sign-in identity belongs to an employee who has left.

C:

A copied personal refresh token remains dependent on a departed employee and introduces credential exposure and lifecycle risk.

Question 40

A contractor needs read-only access to experiment metrics but no ability to download sensitive datasets. What should the administrator evaluate?

  1. Use a dedicated federated or managed workload identity with minimal Azure RBAC.
  2. Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner.
  3. Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access.
  4. Give the endpoint deployment's runtime identity the necessary feature-store data permissions.

Explanation

Correct Answer

C:

Workspace metadata permission does not automatically define every storage grant; separate resource policies protect restricted data.

Incorrect Answers

A:

Use a dedicated federated or managed workload identity with minimal Azure RBAC. It addresses another case: A GitHub deployment workflow needs access to Azure ML without storing an administrator password. For that case, the distinguishing justification or acceptance check is: Workload identity avoids persistent administrator secrets and can be constrained to the specific resources it deploys. The present case instead states: A contractor needs read-only access to experiment metrics but no ability to download sensitive datasets. It calls for Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access. Showing success for the other case would not establish: Workspace metadata permission does not automatically define every storage grant; separate resource policies protect restricted data.

B:

The alternative, Grant a task-scoped Azure ML role at workspace scope instead of subscription Owner, belongs to a different technical condition: A junior operator needs to submit ML training jobs but must not change workspace networking or grant user roles. There the source-specific mechanism or proof is: Least-privilege workspace RBAC permits job operations without broad control over security or billing resources. The condition asked about here is: A contractor needs read-only access to experiment metrics but no ability to download sensitive datasets. A suitable correction is Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access; the targeted justification or result is: Workspace metadata permission does not automatically define every storage grant; separate resource policies protect restricted data.

D:

Give the endpoint deployment's runtime identity the necessary feature-store data permissions. That operation would be relevant if: A production endpoint can fetch its model but cannot read a secured feature store at inference time. Its corresponding reason or verification is: Production inference uses its execution identity; author privileges do not automatically travel to the deployed container. However the problem in this question is: A contractor needs read-only access to experiment metrics but no ability to download sensitive datasets. The requirement instead calls for Use narrowly scoped roles and separate storage data-plane permissions for experiment versus dataset access; the question-specific check is: Workspace metadata permission does not automatically define every storage grant; separate resource policies protect restricted data.

Operationalizing Machine Learning and Generative AI Solutions (AI-300) Practice Tests