Practice Free AI-200 Exam Online Questions
You need to implement responsible AI practices in an Azure ML pipeline.
Which two tools from the Responsible AI dashboard help identify model fairness issues and data issues respectively?
- A . Data explorer /Data analysis component
- B . Error analysis component
- C . Counterfactual analysis component
- D . Interpretability (feature importance) component
- E . Fairness assessment component
You are building an AI agent that must respond to user queries in under 500ms. You notice that the ‘Semantic Ranker’ is adding too much latency.
What is the best optimization strategy?
- A . Increase the Search Tier
- B . Disable the indexer
- C . Use a basic keyword search only
- D . Reduce the ‘top’ number of results sent to the ranker
You are designing the chunking strategy for a RAG system with technical manuals containing tables, images, and long paragraphs.
Which chunking approach is most appropriate?
- A . Recursive character text splitter
- B . Layout-aware chunking preserving table structure
- C . Semantic chunking based on paragraph and section boundaries
- D . Fixed-size token chunking (512 tokens)
- E . Sentence-window chunking
You are developing a customer support bot using Azure OpenAI. You need to ensure the bot does not generate responses containing hate speech or self-harm content without building a custom model.
Which solution should you implement?
- A . Implement a regex-based client-side filter
- B . Increase the temperature setting of the model
- C . Enable Azure AI Content Safety filters
- D . Use a fine-tuned GPT-3.5 model
You are deploying a custom container to Azure Container Apps. You want the app to scale up when the CPU usage across all replicas exceeds 70%.
Which type of scaling rule is this?
- A . HTTP Scaler
- B . Memory Scaler
- C . Custom KEDA scaler
- D . Azure Resource Scaler (CPU)
An AI workload on AKS is reporting high memory usage. You want to see which specific pod is consuming the most resources right now.
Which command provides this real-time resource usage?
- A . kubectl top pods
- B . kubectl logs
- C . kubectl describe pods
- D . kubectl get pods
You are building an AI agentic workflow using Azure Logic Apps. You want to create a loop that processes 100 customer feedback entries in parallel to speed up sentiment analysis.
Which setting should you adjust in the ‘For Each’ loop?
- A . Concurrency Control
- B . Retry Policy
- C . Timeout
- D . Tracking
You are implementing responsible AI measurement for an Azure OpenAI solution. You need to measure whether the model responds differently to equivalent questions about different demographic groups.
Which evaluation methodology detects this?
- A . Azure ML Fairness assessment with demographic cohorts
- B . Differential testing ‘” compare response quality metrics across demographic groups
- C . Counterfactual fairness evaluation ‘” test same question with different demographic identifiers
- D . Red-teaming with demographic-specific adversarial prompts
- E . Human evaluation panel with diverse reviewers
You are building an Azure AI solution that must log every model interaction in an immutable, queryable format for SOC 2 compliance. Log entries must be retained for 7 years.
Which Azure architecture meets these requirements?
- A . Azure Event Hub → Azure Stream Analytics → Azure SQL Database
- B . Azure Monitor Diagnostic Settings → Log Analytics Workspace with 7-year retention
- C . Azure Monitor Diagnostic Settings → Archive to Azure Storage with lifecycle management
- D . Azure Monitor Diagnostic Settings → Azure Storage Account with immutable WORM policy and 7-year retention lifecycle
- E . Azure Application Insights with custom event logging
An organization uses Azure AI Services with virtual network integration. A developer reports that calls from outside the VNet are failing.
Which configuration allows specific external IP addresses while keeping VNet access?
- A . Add an outbound rule to the Network Security Group
- B . Set network access to ‘Selected networks’ and add both VNet and IP rules
- C . Deploy an API gateway outside the VNet to proxy requests
- D . Disable VNet integration and use private endpoints instead
- E . Configure firewall rules on the Azure AI Services resource to allow specific IP ranges