Hard — AI-103 Practice Quiz

Question 1

You are building a RAG ingestion pipeline for a library of engineering white papers that contain descriptive charts and diagrams embedded within the text. Your grounding requirement is that when an agent answers a question about a chart, the retrieved chunk must include a natural-language description of what the chart depicts (trends, axes, labeled regions) rather than just the surrounding paragraph text, so the model can reason about the visual without receiving the raw image at query time. You are using Azure AI Content Understanding to produce the grounded representation. Which configuration produces the required output for downstream reasoning?

Accepted Answer

Configure a document analyzer that emits markdown output and enable generation of natural-language descriptions for embedded figures, so each diagram is represented as descriptive text inline with the surrounding content.

Answer

Use a layout-only analyzer that returns bounding-box coordinates and reading order for each figure region, then store the coordinates alongside the text chunks.

Answer

Run an OCR skill over the pages to extract any text printed inside the charts, and rely on that extracted text as the chart's grounded representation.

Answer

Extract each figure as a separate binary image projection and attach the raw image bytes to the chunk so the retriever can pass them to the model at query time.

Question 2

A financial services team uses Azure AI Content Understanding to extract line items from vendor invoices. They need the extracted vendor names and account codes validated against an authoritative master list of approved vendors so that the analyzer can flag entries that do not match a known record. Which capability should they configure to achieve this validation during extraction?

Accepted Answer

Use Content Understanding pro mode with a reference data source (knowledge base) linked to the analyzer so extracted fields are cross-referenced against the approved vendor list

Answer

Increase the analyzer's confidence threshold so only high-confidence vendor extractions are returned

Answer

Add a custom OCR skill to the ingestion skillset to improve character recognition of vendor names

Answer

Enable markdown output on the analyzer so the vendor names appear as headings for easier manual review

Question 3

You are building a semiautonomous Azure AI Foundry agent for an e-commerce operations team. The agent uses a custom function tool named submit_refund that calls an internal payments API to issue customer refunds. During load testing you notice that transient network timeouts occasionally cause the agent's runtime to retry the tool call, resulting in some customers being refunded twice. You must prevent duplicate refunds without disabling automatic retries. What is the most effective approach?

Accepted Answer

Add an idempotency key parameter to the submit_refund tool schema and have the payments API deduplicate requests that share the same key

Answer

Lower the agent's temperature parameter to 0 so the model produces the same tool arguments on every call

Answer

Set tool_choice to 'none' so the model stops invoking submit_refund automatically

Answer

Increase the function tool's max_tokens so the completion is never truncated mid-call

Question 4

You are building an Azure AI Foundry agent that must call your company's internal Orders REST API, which is protected by OAuth 2.0 client-credentials flow against Microsoft Entra ID. You register the API as an OpenAPI-specified tool for the agent. When configuring the tool, you must ensure the agent can authenticate to the API without embedding long-lived secrets in the tool definition and while respecting the API's token audience requirements. Which authentication configuration should you use for the OpenAPI tool?

Accepted Answer

Configure the OpenAPI tool with a Managed Identity authentication connection so the agent acquires Entra ID tokens scoped to the API's audience at call time.

Answer

Configure the OpenAPI tool with anonymous authentication and rely on the agent's network isolation to protect the API.

Answer

Configure the OpenAPI tool with an API key connection and store the client secret directly in the OpenAPI specification's securitySchemes section.

Answer

Configure the OpenAPI tool to pass the end user's chat session token directly to the Orders API on each request.

Question 5

A financial services team is deploying an Azure AI Foundry agent that can execute stock trades through a registered tool. Regulatory policy requires that any trade exceeding $10,000 be reviewed by a licensed broker before execution, while smaller trades and read-only balance inquiries may proceed automatically to maintain responsiveness. Which oversight configuration best satisfies these requirements?

Accepted Answer

Configure a conditional human-in-the-loop approval mode that intercepts tool calls only when the trade amount exceeds the $10,000 threshold, allowing other actions to run autonomously

Answer

Set the agent to fully autonomous mode and rely on the content safety filter to block high-value trades

Answer

Require human approval for every tool invocation the agent attempts, regardless of the action or amount

Answer

Disable the trading tool entirely and have the agent only return recommendations that brokers manually execute

Question 6

You are building an Azure AI Foundry agent for a travel-booking assistant. The agent has three registered tools: get_user_preferences (reads a profile store), search_flights (queries an airline API using the retrieved preferences), and check_weather (queries a public weather API for the destination). During a single turn the agent must produce a recommendation that uses flight results filtered by user preferences, while weather information is independent. You want to minimize end-to-end latency without producing incorrect results. How should you design the tool-calling behavior for this turn?

Accepted Answer

Invoke get_user_preferences first, then invoke search_flights and check_weather in parallel once preferences are available.

Answer

Invoke all three tools in parallel immediately, then discard any results that depend on data that arrived late.

Answer

Invoke all three tools strictly sequentially in registration order to guarantee deterministic output.

Answer

Invoke check_weather first because it has no dependencies, then run get_user_preferences and search_flights in parallel.

Question 7

A financial services company is designing an Azure AI Foundry solution that must serve customers in the European Union. Legal compliance requires that all customer data used for inference and all model processing remain within EU borders. The team also needs guaranteed capacity for a GPT-4o model deployment during predictable business-hour peaks, with no cross-region data transfer under any circumstances. Which combination of choices best satisfies these requirements?

Accepted Answer

Deploy the model to an EU-based Azure region using a Data Zone Provisioned deployment scoped to the EU data zone.

Answer

Deploy the model to an EU-based Azure region using a Global Standard deployment to maximize availability across the global model fleet.

Answer

Deploy the model to any available region using a Standard (pay-as-you-go) deployment and rely on customer-managed keys for compliance.

Answer

Deploy the model to an EU-based region using a Global Provisioned deployment to guarantee throughput while keeping the resource in the EU.

Question 8

A financial services team builds a loan pre-qualification assistant in Azure AI Foundry. Certain decisions—such as rejecting applicants below a legally mandated minimum credit score—must be enforced with 100% consistency and be fully auditable, while free-form explanation and customer Q&A should be handled conversationally. The team wants to combine an LLM with a deterministic rules engine. Which orchestration approach best satisfies both the compliance and conversational requirements?

Accepted Answer

Evaluate hard eligibility rules in the deterministic rules engine first, then use the LLM only to generate natural-language explanations and answer follow-up questions about the rules-based decision.

Answer

Route all requests through the LLM and add a detailed system prompt instructing it to always enforce the minimum credit score rule.

Answer

Fine-tune the LLM on historical approval data so it internalizes the credit score threshold and returns compliant decisions directly.

Answer

Lower the model temperature to 0 so the LLM produces deterministic outputs that reliably enforce the credit score rule.

Question 9

A financial services team is building a customer support agent in Azure AI Foundry. The agent must answer questions using a knowledge base of 40,000 policy documents that are updated weekly, and it must consistently adopt the company's formal brand tone and a specific structured response format. The team wants to minimize ongoing maintenance overhead while keeping answers accurate against the latest documents. Which approach should they adopt?

Accepted Answer

Use retrieval-augmented generation (RAG) over an indexed vector store for the documents, and shape tone and format through the system prompt and few-shot examples.

Answer

Fine-tune a base model on both the 40,000 documents and the brand tone examples, then redeploy the fine-tuned model each week.

Answer

Fine-tune the model on the brand tone and format examples, and load all 40,000 documents into the system prompt as context on every request.

Answer

Fine-tune the model weekly on the tone examples and rely on the model's pretrained general knowledge to answer policy questions.

Question 10

You are building a customer support solution in Azure AI Foundry that requires several specialized agents (billing, technical, and escalation) to collaborate. The orchestration logic is complex: agents must dynamically decide which peer to invoke next based on runtime conversation state, and the flow cannot be predetermined as a fixed sequence. You want to author this coordination logic in code using an open-source multi-agent framework that Microsoft has consolidated its agentic development around, with native integration into Foundry. Which approach should you choose?

Accepted Answer

Use the Microsoft Agent Framework (which unifies Semantic Kernel and AutoGen) to code the multi-agent orchestration with dynamic, graph-based agent-to-agent handoffs

Answer

Deploy each agent as an isolated managed online endpoint and hard-code a fixed round-robin call order in the calling application

Answer

Configure a single Foundry agent with all three tool schemas and rely on max_tokens tuning to separate the concerns

Answer

Use a static DAG workflow with fixed edges to route every request through billing, then technical, then escalation in that exact order

Question 11

You are orchestrating a multi-agent solution in Azure AI Foundry for an insurance claims workflow. A 'Triage' agent classifies incoming claims and must pass the extracted claim number, policy ID, and damage category to a downstream 'Assessment' agent that runs in a separate thread. The Assessment agent frequently fails because it receives a natural-language summary rather than the exact structured field values, causing lookup errors. What is the most reliable way to transfer this data between the agents?

Accepted Answer

Define a structured output schema for the Triage agent and pass the validated JSON object as explicit input parameters to the Assessment agent's invocation.

Answer

Increase the Assessment agent's max_tokens so it can parse the longer natural-language summary from the Triage agent.

Answer

Lower the temperature on the Triage agent to 0 so its natural-language summary becomes deterministic and easier to parse.

Answer

Store the summary in each agent's conversation memory and rely on thread continuity to share the values.

Question 12

You are building a RAG ingestion pipeline in Azure AI Search for a corpus of long technical manuals (200+ pages each). Retrieval must return small, focused passages for precise grounding, but the agent needs enough surrounding context to reason accurately. During ingestion you must split each document into fine-grained chunks for embedding while still being able to hydrate the full parent section into the agent prompt at query time. Which indexing approach in the skillset should you configure?

Accepted Answer

Configure an index projection that maps each small text chunk to a child index document while preserving a reference to the parent document via a parent key.

Answer

Use the Split skill in pages mode with a large page length so each chunk equals a full section, then embed each large chunk directly.

Answer

Store the entire document text in a single searchable field and rely on semantic ranker to extract the relevant passage at query time.

Answer

Enable integrated vectorization on the whole document field so the embedding model automatically produces one vector per manual.

Question 13

You are building an Azure AI Search RAG ingestion pipeline for a library of long technical manuals. A skillset splits each manual into text chunks and generates vector embeddings per chunk. You need the index to store one document per chunk for fine-grained retrieval, while also preserving the parent manual's metadata (title, product line, revision date) so it can be returned as grounding context. Which indexer configuration achieves this?

Accepted Answer

Configure index projections in the skillset, defining a selector that maps each chunk to its own index document and includes the parent-level metadata fields.

Answer

Set the indexer's parsingMode to jsonArray so each chunk in the split array becomes a separate search document automatically.

Answer

Enable an output field mapping that flattens the chunk array into a single collection field on one document per manual.

Answer

Use a knowledge store table projection to persist chunks, then point the index at the knowledge store as its data source.