AI-103 mock exam

Question 1, Plan and manage an Azure AI solutionYour app calls a Foundry model with an API key stored in app settings. Security wants no secrets in configuration. What should you use?

• AA shared admin account whose password is stored in Key Vault
• BA managed identity with a role assignment on the resource
• CThe same key moved into an environment variable on the host
• DA longer API key that a scheduled job rotates every month

Question 2, Plan and manage an Azure AI solutionThrottling errors appear only during peak hours, while average usage is well under your limit. What do you check first?

• AThe vector index dimensions and how often the index is refreshed
• BThe content filter severity set on the deployment
• CThe model's temperature and top-p settings for those requests
• DThe deployment's rate limits and quota against peak traffic

Question 3, Plan and manage an Azure AI solutionAn agent can call tools that change production data. What is the right governance control for those tools?

• AA lower model temperature so the agent takes fewer risky actions
• BOversight modes and approval steps on those tool calls
• CA larger context window so the agent remembers the rules for longer
• DLeaving the tools out of the documentation that users can see

Question 4, Implement generative AI and agentic solutionsUsers search a product catalog by exact part numbers and by vague descriptions. Which retrieval approach handles both?

• APure vector search with a better embedding model
• BHybrid search combining keyword and vector
• CKeyword search with a larger synonym list
• DSorting results by upload date and popularity

Question 5, Implement generative AI and agentic solutionsYour RAG answers sound confident but cite content that does not exist in the source. Which evaluation targets this?

• AGroundedness evaluation against the retrieved content
• BLatency percentile analysis across the retrieval and generation steps
• CToken count analysis per request and per user
• DFluency scoring on a sample of production answers

Question 6, Implement generative AI and agentic solutionsWhich set of things must you define when building a Foundry agent?

• AThe GPU model, the driver version and the CUDA runtime the agent needs
• BThe hash of the training data and the fine-tuning parameters
• CThe virtual network's subnet size and its peering configuration
• DThe agent's role, its goals, how it tracks conversation, and its tool schemas

Question 7, Implement computer vision solutionsAn attacker hides instructions as tiny text inside an uploaded image, and the model follows them. What is this called?

• AModel drift caused by poisoned inputs
• BTraining data skew
• CToken exhaustion through a crafted image
• DIndirect prompt injection

Question 8, Implement text analysis solutionsA contact center needs live call transcripts translated for agents, and product names keep being misheard. What do you implement?

• AA nightly text analysis job over stored transcripts
• BA speech pipeline with a custom speech model and translation
• CA retrieval index built directly over the raw call audio
• DAn image generation flow that turns each call into a card

Question 9, Implement information extraction solutionsYou need to feed mixed scanned contracts into an agent as clean, structured markdown. Which service fits?

• AAzure Blob Storage with lifecycle rules
• BAzure Speech in Foundry Tools
• CAzure AI Language key phrase extraction
• DAzure Content Understanding