AI-102 Hard Questions — 25 Difficult Questions

Question 1

A developer is using Azure OpenAI to generate code snippets. The developer needs to ensure that the generated code does not contain security vulnerabilities. Which TWO actions should the developer take? (Choose two.)

Accepted Answer

Include examples of secure coding practices in the prompt.. Option A is correct because including examples of secure coding practices in the prompt (few-shot prompting) directly guides the model to generate code that follows those patterns. This technique leverages in-context learning, where the model uses the provided examples to shape its output, reducing the likelihood of producing insecure code without requiring fine-tuning or external filtering.

Answer

Set the max_tokens parameter to a high value to allow longer outputs.

Answer

Fine-tune the model on a dataset containing examples of insecure code.

Answer

Add a system message that instructs the model to never generate insecure code.

Question 2

A security company uses Azure Video Analyzer on IoT Edge to detect intrusions. The edge device has limited compute and network. They need to reduce latency. What should they configure?

Accepted Answer

Set a high minimum confidence threshold for detection.. Option B is correct because setting a high minimum confidence threshold for detection reduces the number of false positives and the volume of events that need to be processed and transmitted. This directly lowers the computational load on the edge device and reduces network bandwidth usage, thereby decreasing latency for actionable intrusion alerts.

Answer

Increase the video resolution sent to the cloud.

Answer

Enable cloud-based processing for all frames.

Answer

Use multiple AI models simultaneously.

Question 3

A company uses Azure Face API to detect faces in a crowd. They need to comply with GDPR and delete face data after 30 days. What should they implement?

Accepted Answer

Use the Face API Delete operation to remove stored face IDs.. Option C is correct because the Azure Face API stores face data (e.g., face IDs) in a PersonGroup or FaceList, and GDPR compliance requires the ability to delete personal data on demand. The Face API provides a dedicated Delete operation (e.g., FaceList - Delete Face or PersonGroup Person - Delete) to remove stored face IDs and associated face data, enabling the 30-day deletion requirement without recreating the resource.

Answer

Enable data encryption at rest.

Answer

Set a retention policy on the Face API resource.

Answer

Recreate the Face API resource every 30 days.

Question 4

A financial services company is building an agent that uses Azure OpenAI to generate investment advice. The agent must be monitored for toxicity and bias. Which combination of services should the team use to implement content safety monitoring?

Accepted Answer

Azure AI Content Safety and Azure OpenAI content filtering.. Azure AI Content Safety provides built-in models for detecting harmful content such as hate speech, self-harm, and sexual content, while Azure OpenAI content filtering applies configurable severity-level filters (e.g., low, medium, high) to model inputs and outputs. Together, they enable real-time monitoring of toxicity and bias in generated investment advice, meeting compliance requirements for financial services.

Answer

Azure Cognitive Search and Azure AI Language.

Answer

Azure Bot Service and Azure Logic Apps.

Answer

Azure Machine Learning and Azure Functions.

Question 5

An agent uses Azure OpenAI with function calling to perform actions. The agent is not executing functions correctly. Which THREE factors should the team check to diagnose the issue?

Accepted Answer

The token limit is too low, truncating the function definitions.. Option B is correct because if the token limit is too low, the model may not receive the full function definitions, causing it to omit or misinterpret available functions. This truncation prevents the model from correctly selecting or formatting function calls, leading to execution failures.

Answer

The temperature parameter is set too high.

Answer

The model version is outdated.

Question 6

A healthcare provider uses Azure Video Analyzer for Media to extract insights from surgical videos. They need to ensure that no patient health information (PHI) is stored in the transcriptions. What is the best approach?

Accepted Answer

Enable content moderation in Video Analyzer for Media settings.. Option C is correct because enabling content moderation in Azure Video Analyzer for Media automatically redacts or flags sensitive content, including PHI, from transcriptions during the indexing process. This built-in feature ensures that PHI is not stored in the output without requiring additional post-processing or manual deletion.

Answer

Use Azure AI Content Safety to post-process transcriptions.

Answer

Disable indexing for videos containing PHI.

Answer

Use the 'delete' API to remove all transcripts after processing.

Question 7

Which THREE are required to build a custom video indexing solution using Azure Video Analyzer for Media? (Choose three.)

Accepted Answer

An Azure Video Analyzer for Media account.. An Azure Video Analyzer for Media account is required because it is the core service that provides the video indexing capabilities, including AI-powered insights like speech transcription, face detection, and sentiment analysis. Without this account, you cannot access the indexing APIs or manage indexed videos.

Answer

A custom indexing pipeline.

Answer

A custom language model.

Question 8

A company is using Azure OpenAI Service to power a customer support agent. The agent sometimes generates incorrect information when it cannot find an answer in the knowledge base. The team wants to ensure the agent only responds using information from the knowledge base and explicitly states when it does not know the answer. Which configuration should the team use?

Accepted Answer

Use 'use your own data' feature with strict content filtering and set the model to only respond based on retrieved documents.. Option C is correct because the 'use your own data' feature in Azure OpenAI Service allows you to restrict the model to answer only from the retrieved documents, ensuring responses are grounded in the knowledge base. Strict content filtering further prevents the model from generating unverified information, and the explicit setting to respond based solely on retrieved documents directly addresses the requirement to state when it does not know the answer.

Answer

Use a custom model fine-tuned on the knowledge base and disable content filtering.

Answer

Use a system message that says 'If you don't know, say you don't know' and rely on the model's training.

Answer

Use prompt engineering with a system message that instructs the model to only answer from the knowledge base, with no additional filtering.

Question 9

An agent uses Azure AI Language to perform sentiment analysis on customer feedback. The team notices that the sentiment scores are sometimes inaccurate for negative feedback. Which TWO improvements should the team consider?

Accepted Answer

Pre-process the text to handle negations and sarcasm.. Option A is correct because Azure AI Language's pre-built sentiment analysis models can struggle with linguistic nuances like negations (e.g., 'not good') and sarcasm (e.g., 'Great, another delay'). Pre-processing the text to explicitly handle these patterns—such as by expanding contractions or using negation detection—can improve the accuracy of the sentiment scores before they are passed to the model.

Answer

Switch to a different language model without fine-tuning.

Answer

Increase the number of decimal places in the sentiment score.

Answer

Increase the confidence threshold for positive sentiment.

Question 10

A company plans to use Azure Cognitive Search to index a large number of PDF documents stored in Azure Blob Storage. The documents contain sensitive personally identifiable information (PII). The search results should only be accessible to authorized users based on their Azure Active Directory (Azure AD) group membership. Which combination of features should the company implement?

Accepted Answer

Index the documents and use security filters with Azure AD group-based access control.. Option B is correct because Azure Cognitive Search supports security trimming via OAuth 2.0 and Azure AD group-based access control. By storing group identifiers in a search index field and using a security filter, only users whose Azure AD group membership matches the stored group IDs can see the corresponding search results. This ensures that sensitive PII in PDF documents is only returned to authorized users without exposing the documents themselves.

Answer

Generate shared access signatures (SAS) for each document and include them in the search index.

Answer

Use Azure RBAC to assign permissions to users for the search service and configure field-level security.

Answer

Store Azure AD group membership in a field in the search index and configure an indexer to map the field.

Question 11

A company is using Azure Cognitive Service for Language to analyze customer support transcripts. They want to identify custom categories (e.g., 'billing', 'technical support') using a custom text classification model. After training and deploying the model, they receive many false positives for the 'billing' category. What is the best first step to improve model accuracy?

Accepted Answer

Review the training data for the 'billing' category and correct any mislabeled examples.. Option C is correct because false positives for a specific category like 'billing' most often stem from mislabeled or ambiguous training examples in that category. By reviewing and correcting the training data for 'billing', you directly address the root cause of the model's confusion, which is the most effective first step in custom text classification model improvement.

Answer

Add more training data to all categories to improve overall model performance.

Answer

Use a different Azure AI service, such as key phrase extraction, to identify billing-related content.

Answer

Increase the confidence threshold for the 'billing' category to reduce false positives.

Question 12

You are an AI engineer at a healthcare company. The company uses Azure Cognitive Services to process medical records. They have a Computer Vision resource deployed in the East US region. Recently, they implemented a custom vision model for detecting specific anomalies in X-ray images. The model was trained using the Custom Vision portal and exported as a TensorFlow model. They deployed the model to an Azure Kubernetes Service (AKS) cluster using a Docker container. The container runs the model and exposes a REST API endpoint for inference. The endpoint is used by a web application that is also hosted in the same AKS cluster. The web application is experiencing high latency when making inference requests. The latency spikes up to 10 seconds during peak hours. The AKS cluster has autoscaling enabled based on CPU metrics. The container's resource limits are set to 1 CPU core and 2 GB memory. The model's inference time on a single image is approximately 500 ms on the development machine. The team has not changed the model or the application code recently. The number of concurrent users has increased by 50% in the last month. What should you do to reduce inference latency?

Accepted Answer

Increase the CPU limit for the container to 4 cores.. Option D is correct because the inference latency is caused by CPU saturation during peak hours. The container is limited to 1 CPU core, and with a 50% increase in concurrent users, the single core becomes a bottleneck, causing inference times to spike. Increasing the CPU limit to 4 cores allows the model to process multiple requests in parallel, reducing queue wait times and overall latency.

Answer

Increase the memory limit for the container to 8 GB.

Answer

Deploy additional replicas of the container and use a load balancer.

Answer

Optimize the model by quantizing it to reduce inference time.

Question 13

A hospital uses Azure Cognitive Service for Language to extract medical entities from clinical notes. The extraction accuracy for medication names and dosages is low. The engineer needs to improve performance without adding new training data. Which solution should the engineer implement?

Accepted Answer

Use custom entity recognition with a prebuilt healthcare entity component.. Option B is correct because the engineer can use custom entity recognition with a prebuilt healthcare entity component, which leverages the existing Text Analytics for Health model's pre-trained entities (including medication names and dosages) without requiring additional training data. This approach combines the prebuilt healthcare model's high accuracy for medical entities with custom entity recognition to fine-tune extraction for specific clinical notes, improving performance without adding new annotated data.

Answer

Add more training data with annotated entities.

Answer

Retrain the Text Analytics for Health model with additional labeled data.

Answer

Increase the confidence threshold for entity extraction.

Question 14

A company uses Azure OpenAI to generate product descriptions. They notice that the model occasionally produces descriptions that include false claims about product features. The company needs to reduce the frequency of these inaccuracies without changing the training data. Which parameter adjustment would be most effective?

Accepted Answer

Decrease the temperature parameter. Decreasing the temperature parameter reduces the randomness of the model's output, making it more deterministic and less likely to generate creative but factually incorrect statements. This directly addresses the need to reduce false claims without modifying training data, as lower temperature forces the model to rely on its most probable (and typically more accurate) token predictions.

Answer

Increase the top_p parameter

Answer

Increase the max_tokens parameter

Answer

Increase the frequency_penalty parameter

Question 15

A company is designing a solution that uses Azure Cognitive Services for text analytics. The solution must meet the following requirements: 
- Detect sentiment in customer feedback. 
- Extract key phrases from the feedback. 
- Identify the language of the feedback automatically. 
- Ensure that the solution can scale to handle thousands of requests per second. 
Which TWO actions should the company take?

Accepted Answer

Use the Text Analytics API for sentiment analysis and key phrase extraction.. Option B is correct because the Text Analytics API (part of Azure Cognitive Services for Language) provides built-in capabilities for sentiment analysis, key phrase extraction, and language detection within a single API call. This eliminates the need for separate services and directly satisfies the requirements for detecting sentiment, extracting key phrases, and identifying language automatically.

Answer

Deploy a separate Language Detection service for language identification.

Answer

Use the Free pricing tier to reduce costs.

Answer

Use Azure Functions to aggregate results from multiple API calls.

Question 16

A company uses Azure Content Moderator to moderate user-generated content. They need to ensure that content moderation workflows comply with regional regulations. Which TWO actions should they take?

Accepted Answer

Deploy Content Moderator resources in the required geographic regions to meet data residency requirements.. Deploying Azure Content Moderator resources in the required geographic regions ensures that user-generated content is processed and stored within specific data boundaries, directly addressing data residency regulations. This is a fundamental compliance requirement because Azure resources are region-bound, and data does not leave the selected region unless explicitly configured otherwise.

Answer

Use the free tier to reduce costs while meeting compliance needs.

Answer

Enable geo-tagging on the Content Moderator API to automatically apply region-specific moderation.

Answer

Configure the API to automatically reject any content that violates regional laws.

Question 17

An image was submitted to Azure Content Moderator's Image Moderation API. The application uses a threshold of 0.5 for Category1 (adult) to trigger a review. Based on the exhibit, what should the application do with this image?

Accepted Answer

Escalate the image to a human reviewer because Category1 score exceeds the threshold.. The application's threshold for Category1 (adult) is 0.5, and the exhibit shows a Category1 score of 0.6, which exceeds this threshold. Even though ReviewRecommended is false, the application's custom threshold overrides the default recommendation, requiring escalation to a human reviewer for manual judgment. This aligns with Azure Content Moderator's design where you can set your own thresholds to trigger reviews regardless of the API's built-in recommendation.

Answer

Auto-approve the image because ReviewRecommended is false.

Answer

Auto-reject the image because no offensive terms were found.

Answer

Log the inconsistency and continue processing without action.

Question 18

A company uses Azure Document Intelligence to extract data from tax forms. They need to improve accuracy for a specific field. Which TWO actions should they take?

Accepted Answer

Label more examples of the specific field in the training set. Option A is correct because labeling more examples of the specific field in the training set directly provides the custom model with additional ground-truth annotations for that field. This increases the model's ability to learn the variations in handwriting, formatting, and layout for that field, which is the most effective way to improve extraction accuracy for a targeted field in Azure Document Intelligence custom models.

Answer

Increase the batch size in the analysis request

Answer

Reduce the image resolution to 200 DPI

Answer

Use the prebuilt-tax.us model

Question 19

Which THREE components are required to build a custom named entity recognition (NER) model in Azure Cognitive Service for Language?

Accepted Answer

A set of labeled documents for training.. Option C is correct because a custom NER model in Azure Cognitive Service for Language requires a set of labeled documents for training. These labeled documents define the entities and their spans within text, which the model uses to learn patterns for extraction. Without labeled data, the model cannot be trained to recognize custom entities.

Answer

A QnA Maker knowledge base for entity definitions.

Answer

A LUIS application to handle entity extraction.

Question 20

A company uses Azure Cognitive Search to index customer support emails. They need to implement a custom skill that extracts the sentiment of the email body and also identifies the primary product mentioned. The custom skill is a Python function deployed as an Azure Function. They want to ensure the skill can process multiple documents concurrently and handle errors gracefully. Which THREE configurations should they apply?

Accepted Answer

Set the inputs of the skill to include the 'text' field from the enriched document and the 'product' field from a previous skill.. Option B is correct because the custom skill must receive the email body text for sentiment analysis and the product field from a prior skill (e.g., a key phrase extraction or entity recognition skill) to identify the primary product. This ensures the Azure Function has all required inputs to produce the desired outputs (sentiment score and product name).

Answer

Set continueOnError to false to ensure that if the skill fails for one document, the entire indexer run stops.

Answer

Set the context property to '/document/pages/*' to process each page of the email individually.