Microsoft Azure AI Engineer - AI-102

1. 1.A retail company wants to automatically pull the key topics and named organizations out of thousands of customer support emails. Which Microsoft Foundry service should the AI engineer choose?

   • AAzure AI LanguageCorrect answer
   • BAzure AI Vision
   • CAzure AI Speech
   • DAzure AI Search

   Why

   Azure AI Language provides key phrase extraction and named entity recognition as prebuilt capabilities over raw text. Azure AI Vision analyses images, Azure AI Speech handles audio, and Azure AI Search indexes content for retrieval rather than performing the extraction itself.

2. 2.An application running on an Azure virtual machine must call an Azure AI services endpoint without any API key being stored in configuration files or code. What should the AI engineer implement?

   • AA managed identity assigned to the virtual machine with the Cognitive Services User roleCorrect answer
   • BThe subscription key stored in Azure Key Vault and read at startup
   • CThe subscription key stored in an environment variable on the virtual machine
   • DA scheduled job that regenerates the subscription key every 24 hours

   Why

   A managed identity in Microsoft Entra ID lets the VM obtain a token for the AI resource at runtime, so no secret is ever persisted. Key Vault still stores a key, environment variables and app settings still hold the key value, and regenerating keys does not remove the need to store one.

3. 3.A solution uses Azure AI Vision, Azure AI Language and Azure AI Speech. The finance team wants a single line item on the invoice and the developers want one endpoint and one key. What should be provisioned?

   • AA multi-service Azure AI services resourceCorrect answer
   • BThree single-service resources in the same resource group
   • CThree single-service resources in the same subscription and region
   • DA single Azure AI Search resource with three indexers

   Why

   A multi-service Azure AI services resource exposes one endpoint and one key across the included services and bills them together. Separate single-service resources produce separate endpoints, keys and meters, and a resource group or subscription only groups resources, it does not consolidate endpoints.

4. 4.An AI engineer is running an Azure AI services container on an on-premises host. Which values must be supplied when the container starts so that it can run and report usage? (Select TWO.)

   • AApiKeyCorrect answer
   • BBillingCorrect answer
   • CA storage account connection string
   • DA managed identity client ID
   • EAn Azure AI Search index name

   Why

   Azure AI containers require ApiKey and Billing, the endpoint URI of the corresponding Azure resource, so the container can validate and meter usage against that resource, alongside Eula=accept. A storage account connection string, a managed identity client ID and a search index name are not container start parameters.

5. 5.A generative AI chat application must block user prompts that attempt to make the model ignore its system instructions. Which Azure AI Content Safety capability addresses this?

   • APrompt ShieldsCorrect answer
   • BCustom blocklists
   • CGroundedness detection
   • DProtected material detection

   Why

   Prompt Shields detect jailbreak and indirect prompt injection attacks against the model instructions. Blocklists match specific terms, groundedness detection checks whether an answer is supported by source data, and protected material detection identifies copyrighted content in output.

Practice all 65 questions

Frequently asked questions