AI-102 Exam Dumps - Designing and Implementing a Microsoft Azure AI Solution

To present a list of options with images, you use Adaptive Cards, which allow for rich content such as images, buttons, and formatted text in the conversation.

These cards are typically surfaced within the context of a dialog, which controls the flow of the conversation.

Azure functions are unrelated here.

Entities and utterances belong to natural language understanding, not UI presentation.

Correct Answer: B and D

You have a chatbot that uses Azure OpenAI for generative responses and Azure AI Content Safety to filter objectionable content.

The requirement is to optimize the content filter configurations by running tests on sample questions.

The proposed solution: Using Content Safety Studio → Moderate text content feature.

Comprehensive Detailed ExplanationThis solution is correct because:

Content Safety Studio provides tools to test and fine-tune content filters.

The Moderate text content feature allows you to run sample prompts and responses against filters for categories such as sexual, violence, hate, and self-harm.

This is exactly how you optimize and validate configurations before applying them to production workloads.

Correct Answer: Yes

Microsoft Reference:

Content Safety Studio overview

Azure AI Content Safety

Scenario Recap:

Proposed Solution: Deploy service1 with a public endpoint, and configure a Network Security Group (NSG) for vnet1.

Analysis:

NSGs only control inbound/outbound traffic at the VNet/subnet or NIC level.

Configuring NSGs does not create private connectivity to Cognitive Search.

Traffic still flows over the public endpoint.

Answer: B. No

NSGs filter traffic, but they don’t prevent routing via the public internet. Only Private Link does.

Change Domains to General (compact).

Retain the model.

Export the model.

Comprehensive Detailed ExplanationYou have a Custom Vision project that performs object detection. The project currently uses the General domain. You need to export the trained model for offline use (on a disconnected network).

Steps required:

Change Domains to General (compact).

Export is only supported with a Compact domain.

The General (compact) domain is optimized for exporting to devices (e.g., ONNX, TensorFlow, CoreML, etc.).

If you stay in “General” domain, you cannot export.

Retain the model.

After switching domains, you must retrain the model in the new compact domain to generate updated weights.

Export the model.

Once retrained in a compact domain, you can export the model for offline deployment.

Export formats include ONNX, TensorFlow, CoreML, or Docker container, depending on use case.

Create a new classification model → Not needed; you already have an object detection model.

Change the classification type → Incorrect; the problem is with the domain type, not classification type.

Why not the others?

Change Domains to General (compact).

Retain the model.

Export the model.

Correct Answer Order:

Export a model from Custom Vision

Domains in Custom Vision

Microsoft References

A user User1 must be able to upload datasets and fine-tune existing models in Azure OpenAI.

The role must follow principle of least privilege.

Role analysis:

A. Cognitive Services Contributor: Too broad; allows management of all Cognitive Services resources, not least privilege.

B. Contributor: Overly permissive; allows full management rights on the entire resource, not just OpenAI.

C. Cognitive Services OpenAI User: Grants permissions only to use models (query endpoints). Does not allow fine-tuning or dataset upload.

D. Cognitive Services OpenAI Contributor: Correct role. This provides the ability to manage OpenAI resources, including uploading training data and fine-tuning models.

Correct Answer: D. Cognitive Services OpenAI Contributor

Microsoft Reference:

Azure RBAC roles for Azure OpenAI

Built-in roles: Cognitive Services OpenAI Contributor

Generate alt textTo automatically suggest descriptive captions for images (usable as alt text), use the Computer Vision Describe Image API. It returns natural-language captions summarizing the image content. Endpoint: /vision/v3.2/describe. Setting maxCandidates=1 gives a single best caption.

Detect inappropriate contentTo block adult/racy/gory images with minimal effort, use Computer Vision Analyze Image with the Adult visual feature. This returns flags (isAdult, isRacy, isGory) and confidence scores. The provided option includes Adult,Description which still returns the needed adult/racy/gore signals, satisfying the requirement.

These choices minimize development effort by using built-in Computer Vision capabilities without training a custom model or wiring up multiple services.

Microsoft Learn – Describe images (Computer Vision): “The Describe Image API returns a list of natural language sentences that describe the image.”https://learn.microsoft.com/azure/ai-services/computer-vision/overview-image-analysis#describe-images

Microsoft Learn – Analyze images (Adult/Racy/Gore detection): “Use the Adult feature to detect adult, racy, and gory content.”https://learn.microsoft.com/azure/ai-services/computer-vision/overview-image-analysis#adult-racy-and-gory-content-detection

References

Comprehensive Detailed ExplanationThere are two main types of data workloads: Batch and Streaming.

Batch Workload

Processes data in large groups (batches) at scheduled intervals.

Common when real-time is not required.

Examples: daily sales reports, hourly data refreshes, periodic ETL (Extract, Transform, Load).

Streaming Workload

Processes data in real time or near real time as it arrives.

Used for scenarios where immediate insights or actions are required.

Examples: fraud detection, IoT sensor monitoring, live purchase tracking.

“Data for a product catalog will be loaded every 12 hours to a data warehouse.”

This is scheduled, not continuous. → Batch

“Thousands of data sets per second for online purchases will be loaded into a data warehouse in real time.”

This requires real-time processing. → Streaming

“Updates to inventory data will be loaded to a data warehouse every 1 million transactions.”

Data is grouped and loaded after reaching a threshold (1 million). → Batch

Applying to the scenarios:

Correct Matches:

Product catalog (12 hours) → Batch

Online purchases (real time) → Streaming

Inventory data (1 million transactions) → Batch

Batch processing vs. Stream processing in Azure

Azure Stream Analytics overview

Data processing approaches: Batch vs Real-time

Microsoft References

a tabular form of rows and columns

Relational data is, by definition, data that adheres to the relational model.

In the relational model, data is organized into tables, which consist of rows (records or tuples) and columns (fields or attributes). This is known as the tabular form.

The other options are incorrect:

Unstructured data is the opposite of relational data.

A hierarchical folder structure is used for file organization, not the logical structure of relational data.

Comma-separated value (CSV) files can contain relational data, but the core definition of relational data is the tabular structure, not the file format itself.

Rationale