AI-300 Genaiops Infrastructure — 15 Questions

Question 1

You are managing a multi-region GenAI deployment using Azure AI Foundry. To reduce latency for global users, what routing architecture should you implement?

Accepted Answer

Use Azure Front Door to route traffic to the nearest regional AI Foundry endpoint.. Using Azure Front Door to route traffic to the nearest regional AI Foundry endpoint is correct because Front Door is a global HTTP/HTTPS load balancer that provides low-latency routing based on anycast and performance. It can direct users to the closest healthy endpoint, reducing latency for global users. Traffic Manager with Performance routing is also a option but operates at DNS level and may not provide the same level of acceleration and caching.

Answer

Implement 'Private Link' for every regional endpoint.

Answer

Deploy Azure Traffic Manager with 'Performance' routing.

Answer

Configure 'Global VNet Peering' between model endpoints.

Question 2

Your production agent application requires strict data isolation. You must ensure all requests to Azure OpenAI are routed through a private network. Which configuration is required?

Accepted Answer

Enable 'Public Network Access' set to 'Disabled' on the Azure AI resource.. Setting Public Network Access to Disabled on the Azure AI resource forces all traffic through private endpoints, which are deployed into a virtual network. Combined with a private endpoint and private DNS zone, this ensures requests to Azure OpenAI resolve to a private IP and never traverse the public internet. This is the required configuration for strict data isolation.

Answer

Deploy the model inside a dedicated 'Azure Kubernetes Service' node pool.

Answer

Configure a 'Service Tag' for Azure OpenAI in the Firewall settings.

Answer

Add the 'Virtual Network' to the 'Allowed Subnets' in the AI Service.

Question 3

To optimize token throughput for a high-traffic chatbot, you decide to enable 'Provisioned Throughput' for your Azure OpenAI deployment. What must you consider?

Accepted Answer

Use 'Azure Monitor' to view the 'Tokens Per Minute' metric.. Provisioned Throughput Units (PTUs) provide dedicated capacity, but they require a commitment to a specific throughput level which impacts cost and availability.

Answer

It can only be used with models deployed in the 'East US' region.

Answer

It automatically scales the number of tokens based on CPU usage.

Answer

Check the 'Quota' tab in the Azure Portal.

Answer

Enable 'Diagnostic Settings' for the Azure AI resource.

Answer

It is incompatible with 'Streaming' requests.

Answer

Query the 'ModelUsage' table in Log Analytics.

Question 4

You are deploying a high-throughput Large Language Model on Azure AI Foundry. You need to ensure the underlying infrastructure supports rapid auto-scaling based on request latency. Which configuration should you prioritize in the deployment settings?

Accepted Answer

Configure the 'Scale' settings in the deployment to use a custom metric based on 'Request Latency'.. Azure AI Foundry deployments expose scale settings that can be driven by custom metrics, and configuring autoscale on request latency directly ties capacity to the user-perceived performance signal. When latency rises, the platform adds instances to absorb load, and when it falls, instances are removed to control cost. This is the most direct way to meet the requirement of latency-based rapid auto-scaling.

Answer

Enable 'Load Balancing' in the Azure AI Search resource.

Answer

Increase the 'Quota' of the underlying Azure subscription.

Answer

Set the minimum instance count to 0 in the deployment configuration.

Question 5

You are deploying a custom model in a container. To ensure the model infrastructure is highly available, what is the best practice?

Accepted Answer

Deploy to 'Azure Kubernetes Service' with 'Availability Zones' enabled.. Deploying across multiple Availability Zones ensures that the infrastructure remains operational even if a specific datacenter fails.

Answer

Disable the 'Load Balancer' to save costs.

Answer

Use a 'Standard' tier App Service plan.

Answer

Deploy the container on a single large VM.

Question 6

You have a new model deployment that is incurring high costs. How can you find the top-consuming deployments?

Accepted Answer

Use 'Cost Analysis' in the Azure Portal.. The cost analysis tool in the Azure portal allows you to group costs by 'Resource' or 'Tag' to identify top-consuming deployments.

Answer

Check the 'Activity Log'.

Answer

Check the Azure AI Foundry 'Deployments' page.

Answer

Review the 'Quotas' tab.

Question 7

When configuring a private endpoint for your Azure AI resource, which TWO of the following must be set up?

Accepted Answer

A virtual network with a dedicated subnet.. Option A is correct because a private endpoint is a network interface created inside a subnet of an Azure virtual network, so you must have a VNet with a dedicated subnet (often delegated or reserved for private endpoints) to host it. Option E is correct because private endpoints resolve the Azure AI resource's fully qualified domain name to the private IP via a private DNS zone (e.g., privatelink.cognitiveservices.azure.com), which is required for internal clients to reach the service privately rather than over its public endpoint. Option B is not required because private endpoints use Azure Private Link routing, not a load balancer. Option C is wrong because private endpoint access is governed by private connectivity and NSGs, not a firewall rule allowing all traffic. Option D is wrong because a private endpoint deliberately avoids exposing or requiring a public IP address for the service.

Answer

A load balancer.

Answer

A firewall rule allowing all traffic.

Answer

A public IP address.

Question 8

Your team is deploying a large language model in Azure AI Foundry. You need to ensure that the inference endpoint remains highly available across different regions. Which configuration should you implement?

Accepted Answer

Enable global load balancing using Azure Front Door to route requests to multiple regional endpoint deployments.. Azure AI Foundry supports multi-region endpoints via Azure Front Door or traffic manager integration to distribute requests across regional inference endpoints for high availability.

Answer

Configure a Load Balancer at the VNet level for the model endpoint.

Answer

Set the model deployment to 'High Availability' mode in the Azure AI Foundry portal.

Answer

Replicate the underlying storage account for the model weights across regions.

Question 9

You are configuring an agent orchestration system using Prompt Flow. You need to capture trace data for every step in the agent's reasoning process. Which infrastructure component must be enabled to visualize this in Azure AI Foundry?

Accepted Answer

Enable Azure Monitor diagnostic settings to stream Prompt Flow logs to a Log Analytics workspace.. Enabling the connection to an Azure Log Analytics workspace within the Prompt Flow project settings allows for the storage and visualization of trace data.

Answer

Configure the Prompt Flow runtime to use a Managed Identity with read access to the Blob Storage account.

Answer

Deploy a specialized Azure Function to intercept and log flow run events.

Answer

Increase the timeout threshold in the runtime environment settings.

Question 10

You are securing an agent orchestration infrastructure where agents must access internal data stores. To prevent data exfiltration, you need to restrict all traffic to the Azure AI Foundry project to a private network. What should be configured?

Accepted Answer

Configure a Private Endpoint for the Azure AI project and disable public access.. Configuring a Private Endpoint for the Azure AI project and disabling public access ensures that the Azure AI Foundry project is only reachable via a private IP address within your virtual network, effectively blocking all public internet traffic. This directly satisfies the requirement to restrict all traffic to the project to a private network, preventing data exfiltration. Private Endpoints use Azure Private Link to establish a secure, private connection, and disabling public access removes any exposure to the public internet.

Answer

Enable a Virtual Network Service Endpoint on the storage account only.

Answer

Implement an Azure Firewall to inspect all outbound traffic from the agents.

Answer

Use an API Gateway with IP whitelisting.

Question 11

You are optimizing the cost of your GenAI infrastructure. You have several agents running in Prompt Flow that are idle for large portions of the day. Which runtime configuration should be modified?

Accepted Answer

Set the 'automatic_shutdown' property on the Prompt Flow runtime compute instance.. Setting the 'idle_time_before_shutdown' or similar inactivity policies in the Prompt Flow runtime ensures compute resources are deallocated when not in use.

Answer

Change the VM SKU to a burstable instance type.

Answer

Implement a script to delete the runtime when the flow is finished.

Answer

Set the concurrency limit to zero.

Question 12

Which infrastructure artifact acts as the primary container for organizing and managing your models, deployments, and associated connections in Azure AI Foundry?

Accepted Answer

Azure AI Project. In Azure AI Foundry, an Azure AI Project is the primary container that organizes models, deployments, connections, and associated resources. It provides a workspace-like boundary for managing AI assets and is the top-level resource for Foundry projects. Other options are either broader (resource group) or legacy (ML workspace).

Answer

Azure AI Resource

Answer

Azure Machine Learning Workspace

Answer

Azure Resource Group

Question 13

You are reviewing the infrastructure architecture for a GenAI application that requires strict regulatory compliance. Which TWO of the following configurations are required to ensure data privacy and residency?

Accepted Answer

Disabling the storage of input/output data in the Azure AI service via policy.. Option A is correct because disabling storage of input/output data in the Azure AI service via policy ensures that prompts and completions are not persisted by the service, which is essential for data privacy and regulatory compliance since sensitive data cannot be retained or potentially accessed later. Option B is correct because Private Link creates a private endpoint that keeps traffic between the application and the Azure AI service on the Microsoft backbone network, preventing exposure over the public internet and helping satisfy data residency and privacy requirements. Option C is not relevant because increasing model throughput capacity only affects performance and quota, not data privacy or residency. Option D is incorrect because enabling Public IP access for remote debugging exposes the service to the public internet, directly undermining privacy and compliance. Option E is incorrect because deploying models across multiple global regions can move or replicate data across geographies, which conflicts with strict data residency requirements rather than ensuring them.

Answer

Increasing the model throughput capacity.

Answer

Enabling Public IP access for remote debugging.

Answer

Deploying models across multiple global regions.

Question 14

You are the MLOps engineer for a production GenAI chat application built on Azure AI Foundry. The application calls a model deployment through its online endpoint. During peak hours, users report intermittent 429 responses, but the Azure Monitor metrics for the deployment show token throughput well below the provisioned quota. You need to reduce these throttling errors without redeploying the model. Which two actions should you take? (Choose two.)

Accepted Answer

Distribute requests across multiple deployments of the same model by configuring a load-balancing rule in Azure API Management.. The 429s occur despite low token throughput, which points to request-rate bursts rather than token quota exhaustion. Client-side retry with exponential backoff and jitter smooths those bursts, while distributing traffic across multiple deployments raises the effective request limit. Both actions reduce throttling without redeploying the model, satisfying the scenario constraint.

Answer

Reduce the model's max tokens parameter to a lower value so each request consumes fewer tokens.

Answer

Enable diagnostic logging on the Azure AI Foundry project and wait for the logs to identify the throttled requests.

Answer

Increase the model deployment's tokens-per-minute quota by redeploying the model with a higher capacity SKU.

Question 15

You are deploying a GenAI application that uses Azure AI Foundry and must ensure that prompts and completions are not logged to Microsoft-operated storage for privacy compliance. The application calls a model deployment through the Foundry endpoint. Which configuration should you apply?

Accepted Answer

Disable the abuse monitoring opt-out by setting the appropriate flag on the Azure AI Foundry resource so prompt and completion data is not stored.. The requirement is to prevent prompts and completions from being stored by Microsoft. Azure AI Foundry offers an abuse monitoring opt-out for eligible customers that stops this retention at the resource level. Encryption, private networking, and content filtering address different concerns and do not change whether the service stores the data.

Answer

Enable private endpoints on the Azure AI Foundry resource so that traffic never leaves your virtual network.

Answer

Set the model deployment's content filter to block all categories so that no prompts or completions are processed.

Answer

Configure a customer-managed key on the Azure AI Foundry resource so that all stored data is encrypted with your key.