Free AI-200 Practice Test – 20 Questions with Explanations

A team deploys a containerized embeddings API to AKS. After rollout, the service returns intermittent 503 responses. The team must diagnose the issue by inspecting logs and events before changing manifests, scaling settings, or container images.
Which four actions should you perform in order?
Delete the namespace and redeploy
Identify the pods created by the deployment
Disable all liveness probes
Inspect recent logs for affected pods
Review pod details and recent Kubernetes events
Check service endpoints for ready pod targets

A team deploys a new revision of an AI inference API. The stable revision must keep most production traffic while the candidate revision receives 10 percent for canary validation. The team also wants a stable label URL that routes testers directly to the candidate revision.
Snippet

Evaluate the following statements.
Statement 1: Multiple revision mode is required for simultaneous traffic allocation across active revisions.
Statement 2: The canary label can provide a stable URL that targets the candidate revision.
Statement 3: Setting revisionSuffix is sufficient by itself to send 10 percent of production traffic to the candidate revision.

A containerized summarization worker reads messages from an Azure Service Bus queue named summarize-requests. The worker must scale from zero to ten replicas based on queue depth. The team wants the scale rule to use the app’s system-assigned managed identity instead of a Service Bus connection string stored as a secret.
Snippet

Which option completes the configuration?

Develop containerized solutions on Azure (20-25%)
A development team hosts a containerized AI API in Azure App Service. The container image is stored in Azure Container Registry and the application calls an external embedding service. The application expects EMBEDDING_KEY as an environment variable. Security requires centralized secret rotation and reduced secret copying. Operations wants to update the secret without rebuilding the image. The web app can use a system-assigned managed identity.
Requirements:
– Keep the secret outside the image
– Provide the value through the existing environment variable name
– Support rotation from a centralized secret store
– Avoid placing the secret in deployment scripts
Which configuration should you use? Select one answer.

Develop containerized solutions on Azure (20-25%)
A team stores AI API images in an Azure Container Registry repository named ai-api. Feature branch tags accumulate quickly and increase registry storage usage. You need to schedule cleanup so tags older than 30 days are removed, the five newest matching tags are kept, and untagged manifests are included. The cleanup job must run from Azure Container Registry without a source build context.
Snippet block

Which option completes the configuration?

A multi-tenant RAG API stores chunks in PostgreSQL with structured columns, jsonb metadata, and a vector(1536) embedding column. Retrieval must return semantically similar chunks from the requested tenant and document category before assembling the prompt.
Snippet

Evaluate the following statements.
Statement 1: The query applies tenant and metadata constraints as part of retrieval instead of filtering returned chunks in
application code.
Statement 2: The embedding parameter must match the dimensionality of the embedding column for vector comparison to work
correctly.
Statement 3: The query removes the need for structured, metadata, or vector indexes in a production workload.

A semantic retrieval service stores support article chunks in Azure Cosmos DB for NoSQL. Each item contains tenantId, chunkText, sourceUri, and an embedding array generated by a model that returns 1,536 dimensions. The team must configure the container so the embedding can be stored with the document and used later for vector retrieval.
Snippet

Which option completes the configuration?

An AI application stores user preference documents in Azure Cosmos DB for NoSQL. The application frequently retrieves a single preference document when it already knows both the document id and the partition key value. The current implementation runs a SQL query with WHERE c.id = @id, and RU charges are higher than expected.
Which change should you make?

Develop AI solutions by using Azure data management services
A RAG API stores generated section summaries in Azure Managed Redis. Cached values must be reused when the same document content and model version are used. Cached values must also expire so updated document processing eventually refreshes the response.
Which two actions should you perform? Select TWO answers.

Steps
1. Poll documents by _ts every minute
2. Create or reference a lease container
3. Get a reference to the monitored container
4. Define the delegate that processes change batches
5. Start the change feed processor
6. Store lease documents in the monitored container
An AI enrichment worker must generate embeddings when documents are inserted or updated in an Azure Cosmos DB for NoSQL container. The worker will run on multiple compute instances and must distribute processing safely while tracking progress across partitions. The team chooses the Cosmos DB change feed processor rather than timestamp polling. Which four actions should you perform in order?

Develop AI solutions by using Azure data management services
A RAG API stores embeddings in Azure Database for PostgreSQL. The table contains tenant_id, document_id, embedding, and content. The query filters by tenant and orders by cosine similarity against a query embedding. You need to improve retrieval performance while keeping tenant scoping efficient.
Which two actions should you perform? Select TWO answers.

A containerized API receives document-analysis requests and must enqueue work for asynchronous processing. The API runs with a managed identity and must send messages to a Service Bus queue named analysis-requests. The implementation must avoid storing a connection string and must grant the least data-plane role needed for sending.
Snippet

Which option completes the configuration?

A team needs to route AI moderation requests from a Service Bus topic to a human-review worker. Messages include application properties named category and severity. The worker should receive messages where category is Safety and severity is High.
You need to complete the Azure CLI commands.
Snippet

Which option completes the configuration?

A storage account emits blob-created events when users upload documents for AI extraction. An Azure Function must start the extraction workflow when a matching blob is created. You need to configure the event-driven path, validate delivery, and avoid polling the storage account.
Steps
1. Create an Event Grid subscription from the storage account to the Azure Function endpoint
2. Upload a test blob that matches the expected path
3. Confirm the Function endpoint is deployed and can receive Event Grid events
4. Add a timer trigger to poll the container every minute
5. Configure subject or event-type filtering for the relevant blob-created events
6. Review Function logs for the delivered event and extraction invocation
Which four actions should you perform in order?

Connect to and consume Azure services (20-25%)
An AI document pipeline starts when a PDF is uploaded to Azure Blob Storage. An Azure Function extracts metadata, creates an enrichment request, and records the ingestion status. The enrichment request must be sent to Azure Service Bus for worker processing. The ingestion status must be written to Azure Cosmos DB. Developers want a binding-based design rather than manually creating clients for each output service.
Requirements:
– Start from the blob upload event
– Send a message to Service Bus
– Write ingestion status to Cosmos DB
– Keep service connection details in binding configuration
– Avoid a scheduled polling design
Which implementation should you use? Select one answer.

A team is designing configuration for a containerized AI API. The app needs a model endpoint URL, a feature flag, a retrieval top-k value, and an API key for a private model gateway. The app runs with a system-assigned managed identity.
Evaluate the following statements.
Statement 1: Store the API key directly as an Azure App Configuration key-value so the app has one configuration source.
Statement 2: Store non-secret values in Azure App Configuration and store the API key in Azure Key Vault.
Statement 3: Grant the app identity access to App Configuration data and permission to read the referenced Key Vault secret.

An AI API uses Application Insights and OpenTelemetry instrumentation. Users report intermittent HTTP 500 responses from the /chat/completions endpoint. You need to identify failed requests, correlate them to exceptions and dependency failures, and preserve the operation context across related telemetry.
Snippet

Evaluate the following statements.
Statement 1: Joining on operation_Id helps correlate failed requests with related exceptions and dependency failures.
Statement 2: Filtering success == false or HTTP 5xx result codes narrows the request set to likely failed server-side requests.
Statement 3: This query proves that OpenTelemetry propagation is configured correctly across every downstream Azure service.

Secure, monitor, troubleshoot Azure solutions (20-25%)
A Python AI API runs in Azure Container Apps. The API must send OpenTelemetry traces to Azure Monitor Application Insights. The application uses a RAG workflow and needs a custom span around the answer-generation stage. The Application Insights connection string is stored in an environment variable.
Snippet block

Which option completes the configuration?

Secure, monitor, troubleshoot Azure solutions (20-25%)
You are implementing automated rotation for an external model-provider credential stored in Azure Key Vault. The credential must be replaced before expiry. A rotation function must call the provider, store the replacement as a new Key Vault secret version, and preserve application references to the same secret name.
You need to choose the correct implementation sequence.
Which four actions should you perform in order?
Create or identify the Key Vault secret with expiry metadata
Subscribe the vault event to an Azure Function handler
Grant the function identity permission to set secrets
Write the replacement credential to the same secret name
Store the replacement credential in source control
Rename the secret after every rotation cycle

A team runs an AI enrichment worker in Azure Container Apps. The app uses revisions and event-driven scale rules. After a new revision is activated, requests intermittently fail during startup. The developer must inspect runtime console output and platform-generated messages. The team needs queryable historical data for incident review.
Requirements:
• Inspect container console output.
• Review system-generated platform events.
• Filter records by app and revision.
• Use Azure Monitor log data for history.
Which approach should the team use? Select one answer.