Practice Free AI-200 Exam Online Questions
You are building an AI agent that requires a long-term ‘memory’ store. You choose Azure Database for PostgreSQL.
Which extension must be enabled to allow the database to store and query high-dimensional vector embeddings?
- A . uuid-ossp
- B . postgis
- C . pg_similarity
- D . pgvector
A company uses Azure OpenAI for code generation. They want to prevent the model from generating malicious code.
Which layers of defense should be implemented?
- A . Azure AI Content Safety Prompt Shield
- B . System prompt instructing the model to refuse malicious requests
- C . Azure API Management policy to inspect output
- D . Azure OpenAI built-in content filters for code harm category
- E . Post-processing output validation against known patterns
You need to deploy a containerized Azure AI Services model on-premises for data sovereignty.
Which steps are required to run the Azure AI Language container?
- A . Mount a local volume for model storage
- B . Pull the container image from Microsoft Container Registry (MCR)
- C . Install NVIDIA CUDA drivers for GPU acceleration
- D . Configure the container to run without internet connectivity (disconnected)
- E . Pass billing endpoint URI and API key as environment variables
You are developing an AI-augmented app that stores large amounts of unstructured data in Azure Database for PostgreSQL. You need to perform efficient similarity searches on high-dimensional vectors.
Which extension should you enable?
- A . pgvector
- B . uuid-ossp
- C . PostGIS
- D . pg_cron
An AI application on Azure Container Apps needs to scale horizontally when the ‘tokens-per-second’ metric from an internal gateway exceeds a threshold.
Which scaling mechanism is best suited for this custom metric?
- A . Manual scaling only
- B . CPU/Memory utilization scaling
- C . KEDA Prometheus scaler
- D . Built-in HTTP scaling
A developer is implementing a multi-agent system using Azure AI. Agent 1 extracts data from documents, Agent 2 validates the data, and Agent 3 generates a report. The system must handle failures in Agent 2 gracefully, logging the error and continuing with Agent 3 using partial data.
Which orchestration approach supports this error handling pattern?
- A . LangGraph with error handling nodes and conditional edges
- B . Azure Durable Functions fan-out/fan-in with try/catch in the orchestrator
- C . Azure Service Bus dead-letter queue for failed Agent 2 messages
- D . Semantic Kernel Process Framework with error handling steps
- E . Azure Logic Apps with parallel branches and error handling
You are developing a multi-agent system where agents communicate asynchronously. You need to ensure that if Agent A sends a message to Agent B and Agent B is offline, the message is stored securely until Agent B returns.
Which service is best suited for this?
- A . Azure Event Hubs
- B . Azure Service Bus Queue
- C . Azure Load Balancer
- D . Azure Front Door
You are building a RAG application and need to split a 50-page PDF into smaller, overlapping sections before vectorizing it.
What is this process called?
- A . Normalization
- B . Tokenization
- C . Indexing
- D . Chunking
You are building an agentic workflow that needs to ‘Wait’ for a human to approve an action before the agent proceeds.
Which Logic App pattern should you implement?
- A . Human-in-the-loop (HITL) with a manual approval action
- B . Delayed trigger
- C . Infinite loop
- D . For Each loop
You need to deploy an Azure AI solution that includes a fine-tuned Azure OpenAI model. The solution requires the model to be updated monthly with new training data.
Which workflow automates this?
- A . Azure Machine Learning pipeline with fine-tuning job → evaluation → deployment
- B . Azure Functions scheduled trigger calling fine-tuning API → evaluate → swap deployment
- C . Azure Logic Apps orchestrating fine-tuning API, evaluation, and deployment steps
- D . Azure DevOps pipeline with Azure OpenAI fine-tuning API calls
- E . Azure OpenAI fine-tuning directly in Azure AI Studio with scheduled triggers