Practice Free AI-200 Exam Online Questions
A logistics company utilizes Azure OpenAI Service to generate daily delivery summaries for truck drivers. The operations team reports that the model occasionally hallucinates delivery times and addresses that are missing from the primary database source. You must implement an optimization strategy that minimizes these hallucinations by forcing the model to rely strictly on verified database facts.
Which two techniques should you use? (Select TWO)
- A . Switch the deployment type to an open-source model running on an unmonitored container instance.
- B . Implement a Retrieval-Augmented Generation (RAG) pattern to dynamically inject matching database facts into the prompt system context.
- C . Deploy an Azure AI Vision spatial analysis model to verify the text formatting constraints.
- D . Refine the system prompt instructions to explicitly command the model to output ‘Not Found’ if the answer cannot be derived from the provided context.
- E . Increase the model’s temperature parameter to 1.5 to encourage wider probability exploration.
An AI application needs to process millions of small text chunks for a RAG (Retrieval-Augmented Generation) pipeline. You are using Azure Cosmos DB for NoSQL.
Which indexing policy should you use to optimize for vector search while minimizing write latency?
- A . Manual indexing on every field
- B . Lazy indexing
- C . No indexing
- D . Consistent indexing with excluded non-vector paths
You are deploying a containerized AI model to Azure Container Instances (ACI). You need to ensure the container can read a ‘config.json’ file from an Azure File Share.
What should you configure in the ACI deployment?
- A . A Managed Identity
- B . A Docker environment variable
- C . An HTTP download script
- D . A persistent volume mount
A healthcare provider is deploying an AI triage system using Azure OpenAI Service to summarize emergency room intake notes. The system must adhere to strict regulatory compliance: it must never process text containing explicit descriptions of self-harm, and it must detect and flag any adversarial user prompts trying to bypass the AI’s clinical constraints.
Which two Azure AI Content Safety features should you integrate? (Select TWO)
- A . Configure the Text Analysis API with a strict (Low) severity threshold for the Self-Harm category to block non-compliant input immediately.
- B . Configure a Custom Translator model to obscure sensitive terms before ingestion.
- C . Deploy an Azure AI Vision spatial analysis container to visually inspect text character layouts.
- D . Enable the Prompt Shields feature to detect and intercept jailbreak attempts within the user text.
- E . Implement an Azure AI Search synonym map to dynamically rewrite self-harm descriptions into medical codes.
You are building an AI agent that uses a ‘Tool’ to fetch data from an external SaaS API. You want to ensure the API key is never stored in the code and is automatically updated when changed in Key Vault.
Which pattern should you use?
- A . Environment Variables
- B . Key Vault References in App Service/Functions
- C . Manual deployment updates
- D . Hardcoding in a secret.json
An AI application hosted on Azure Container Apps needs to scale out based on the number of active ‘Jobs’ in an Azure Service Bus queue.
Which component of Container Apps handles this event-driven scaling?
- A . Horizontal Pod Autoscaler
- B . Azure Monitor
- C . KEDA (Kubernetes Event-driven Autoscaling)
- D . Dapr
An IoT solution uses Azure AI Vision Spatial Analysis to detect when more than 10 people are in a restricted zone.
Which components are required for this deployment?
- A . Azure IoT Hub for telemetry ingestion
- B . An IP camera or RTSP video stream
- C . Azure AI Vision resource in Azure
- D . Custom Vision model for person detection
- E . Azure Stack Edge or an NVIDIA GPU-enabled device running the Spatial Analysis container
Your company is building an AI-powered customer support chatbot using the Azure OpenAI Service. During testing, you discover that some users are attempting ‘jailbreak’ attacks by supplying adversarial prompts designed to bypass the safety filters of the model. You need to implement a native defense strategy using Azure AI Content Safety that minimizes latency and specifically targets prompt injection and jailbreak attempts.
Which two features or actions should you configure? (Select TWO)
- A . Enable the Prompt Shields feature for jailbreak detection within Azure AI Content Safety.
- B . Deploy an Azure AI Vision model to analyze the user text as an image.
- C . Configure a text classification model in Azure AI Language to predict jailbreak probability.
- D . Configure custom blocklists within Azure AI Content Safety to flag known adversarial phrases.
- E . Decrease the temperature parameter of the Azure OpenAI model deployment to 0.
You are building an AI agentic workflow using Azure Logic Apps. You need to trigger a workflow only when a new ‘Critical’ level alert is logged in Azure Monitor.
Which connector should you use?
- A . HTTP Webhook
- B . Azure Monitor Logs (KQL trigger)
- C . Azure Event Grid
- D . Office 365 Outlook
You are developing a Retrieval-Augmented Generation (RAG) solution using Azure AI Search as your vector database. The application needs to perform vector search against an index containing document chunks. Your vector field is configured using the Hierarchical Navigable Small World (HNSW) algorithm. You want to adjust the configuration to prioritize high search accuracy (recall) over indexing speed and memory utilization during graph construction.
Which two HNSW parameters should you increase? (Select TWO)
- A . dimensions
- B . metric
- C . m
- D . exhaustiveKnn
- E . efConstruction