Skip to content

AI-200 Developing AI Cloud Solutions on Azure Practice Questions

Prepare for AI-200 with more than an answer.

150 questions in the full set12 sample questionsUpdated Oct 3, 2026
Exam fee
$165 USD
Level
Associate (Intermediate)
Valid for
1 year
Domains covered on the exam 4
  1. Develop containerized solutions on Azure22.5%
  2. Develop AI solutions by using Azure data management services27.5%
  3. Connect to and consume Azure services22.5%
  4. Secure, monitor, troubleshoot Azure solutions22.5%
  1. 1

    A lead AI engineer creates a new container in Azure Cosmos DB for NoSQL with a vector index policy using quantizedFlat to store 1,536-dimensional vectors. During early development, the developer seeds 450 document vectors into the container and runs semantic similarity queries using VectorDistance(c.embedding, @queryVector). How does Cosmos DB execute this query given the current data volume?

    Show answer details

    Correct answer: D

    In Azure Cosmos DB for NoSQL, the quantizedFlat and diskANN vector index types require a minimum of 1,000 vectors to train the vector quantization dictionary. If the container contains fewer than 1,000 vectors, the indexing engine automatically falls back to full brute-force scans (acting like a flat index) without failing the queries.

  2. 2

    True or False: In Azure Cosmos DB for NoSQL, you can alter the vectorEmbeddings policy of an existing container to change the vector distance function from cosine to dotproduct without creating a new container.

    Show answer details

    Correct answer: B

    Vector embedding policies and vector indexes in Azure Cosmos DB for NoSQL are immutable once the container has been created. If you need to modify the path, data type, dimension count, or distance function (cosine, dotproduct, or euclidean), you must create a new container with the updated policy and migrate your data.

  3. 3

    A solution architect is designing a real-time embedding generation pipeline using Azure Cosmos DB for NoSQL. When documents are inserted or modified in the CustomerFeedback container, a Change Feed Processor (CFP) worker service must compute new embeddings and write them back. Another independent audit service also needs to read the change feed of CustomerFeedback from the beginning. Both services share a single lease container named leases. What configuration must be applied to ensure the two worker services operate independently without conflicting over partition leases?

    flowchart TD SourceContainer[(Monitored Container: CustomerFeedback)] --> CFP_AI[AI Embedding Worker Unit] SourceContainer --> CFP_Audit[Audit Service Worker Unit] CFP_AI --> LeaseContainer[(Shared Lease Container: leases)] CFP_Audit --> LeaseContainer
    Show answer details

    Correct answer: C

    When multiple independent deployment units (applications or microservices) share the same lease container to monitor the same source container, each deployment unit must specify a distinct leasePrefix. The leasePrefix scopes lease document IDs in the lease container so that each service maintains its own independent checkpoints and partition locks without interfering with each other.

  4. 4

    A Python developer is building an asynchronous ingestion service for an Azure Cosmos DB for NoSQL database. The service must monitor a container for newly inserted knowledge articles to trigger an external Azure OpenAI embedding workflow. The developer searches for the high-level ChangeFeedProcessorBuilder class in the azure-cosmos Python SDK but cannot locate it. What architectural reality in Azure Cosmos DB explains this, and what is the recommended Python approach?

    Show answer details

    Correct answer: D

    In Azure Cosmos DB, the push-based Change Feed Processor (CFP) with automated lease management and distributed partition balancing is provided exclusively in the .NET and Java SDKs. Python and Node.js applications must use the change feed pull model (query_items_change_feed / change feed iterator), which requires the application to handle continuation tokens and checkpoints explicitly.

  5. 5

    A machine learning engineering team is deploying Python-based backend inference services to Azure. The team wants to build container images directly from application source code within Azure Container Registry (ACR) without authoring or maintaining a Dockerfile, and without running a local Docker daemon. Which Azure CLI command should the team execute?

    Show answer details

    Correct answer: C

    The az acr pack build command leverages Cloud Native Buildpacks directly within Azure Container Registry to compile application source code into a container image without requiring a Dockerfile. In contrast, az acr build performs a quick task build that requires a Dockerfile. az acr run executes commands inside an existing container image in ACR Tasks, and az acr import is used to import already-built images from external container registries.

  6. 6

    A cloud architect is designing an automated container build pipeline using Azure Container Registry (ACR) Tasks for an enterprise RAG application. The solution must automatically recompile and tag the backend container image whenever the underlying Ubuntu-based runtime base image is patched by the vendor, and must also execute a multi-step task defined in a YAML configuration file. Which TWO capabilities should the architect configure? (Select TWO)

    Show answer details

    Correct answer: B, E

    Multi-step tasks in ACR Tasks are declared using YAML files and support built-in run variables such as {{.Run.ID}}, {{.Run.Registry}}, and {{.Values}} to parameterize and tag image outputs.

    ACR Tasks includes native support for base image update triggers (--base-image-trigger-enabled true), which automatically trigger application image builds when an upstream OS or framework base image is updated in the registry. Multi-step tasks in ACR Tasks are declared using YAML files and support built-in run variables such as {{.Run.ID}}, {{.Run.Registry}}, and {{.Values}} to parameterize and tag image outputs. Polling via cron or deploying custom container instances introduces unnecessary complexity when native ACR triggers exist.

Create an account to continue.