AI-102 EXAM QUESTIONS 2026 MICROSOFT AZURE AI 102 CERTIFICATION COURSE PART-2

This video, part of the AI 102 series, revisits previous questions before diving into new topics. A quick review covers essential concepts: * **Computer Vision Client Library for Image Description and Visual Features:** `d.description` * **Computer Vision Client Library for Image Text and Visual Features:** `d.text` * **Reading a File from a Local System:** `file.openread` * **Azure Service for Analyzing Visual Content (Images/Videos) to Categorize:** Azure AI Vision * **Prediction by a Machine Learning Model:** Probabilistic values based on correlations in training data. * **Protecting Account Keys for Azure AI Services:** Azure Key Vault * **Indexing Text Documents for Mobile App Search:** Azure AI Search * **Consuming an Azure Machine Learning Model in an Application:** Publish the model using Azure Machine Learning, then consume via REST API. * **Configuring Diagnostic Logging for Azure AI Services:** Enable diagnostic settings, select log type, configure log destination. * **Validating a Bot Framework Bot Locally:** Run the Bot Framework Emulator. The discussion then transitions to new questions, starting with Azure services for understanding human language, sentiment analysis, and language translation. To grasp these capabilities, it's crucial to understand key elements for language model training: utterances, entities, and intents. * **Utterance:** The user input that the model needs to interpret (e.g., "Turn the lights on"). * **Entity:** The word or phrase that is the focus of an utterance (e.g., "lights" in "Turn the lights on"). * **Intent:** The action or task to be executed based on the entities (e.g., "turn on" in "Turn the lights on"). For example, given the input "I want to book a flight from Jaipur to Kota on 22nd January": * The entire sentence is the **utterance**. * The **intent** is "book a flight," with a high confidence level (e.g., 91.8%). * The **entities** are "Jaipur" (from City), "Kota" (to City), and "22nd January" (flight date), all with high confidence. This demonstrates how a model interprets natural human language. **Sentiment Analysis** is another key capability. If a user states, "The food and service were unacceptable," the model can interpret this as a negative sentiment (e.g., 99% sure). It identifies "food" and "service" as targets and "unacceptable" as the assessment, indicating a negative sentiment for both. **Language Translation** is a straightforward concept, enabling translation between languages. Which Azure service provides these capabilities? * **Azure AI Document Intelligence:** Extracts text, key-value pairs, tables, and structures from documents. Not the answer. * **Azure AI Decision:** Related to decision-making APIs, like content personalization and recommendations. While it utilizes NLP, it's not the primary service for general language understanding. * **Bot Services:** Used for building and deploying bots. Not the answer. * **Azure AI Language:** This is the correct answer. It's an overarching branch of Azure services that provides natural language features to understand and analyze text, encompassing human language understanding, sentiment analysis, and language translation. Next question: **Configuring a web app to connect with an Azure OpenAI model using the Azure OpenAI SDK.** The model is named "model one." What information is required? The name "model one" refers to the *deployment name*, not the actual OpenAI model name (e.g., GPT-3, GPT-4). To establish a connection, you need: * **Deployment Name:** The name assigned to the deployed model (e.g., "model one"). * **Key:** The API key for authentication and authorization. * **Endpoint:** The URL representing where to connect, found in the Azure portal's overview section. Therefore, the correct option includes **deployment name, key, and endpoint**. Model type or actual model name (e.g., GPT-4) is not directly required for the connection itself. Next question: **Designing a system for personalized recommendations.** Which service is most appropriate? * **Azure Functions:** Serverless compute service, not for recommendations. * **Azure AI Personalizer:** This is the correct answer. It's specifically designed to provide personalized content and recommendations, utilizing reinforcement learning. Examples include Netflix recommendations, YouTube suggestions, e-commerce product recommendations, content design (ad placement), and communication (notification timing). * **Azure Machine Learning:** A general platform for ML tasks, not specific to personalization. * **Azure Cognitive Search:** A search engine service, not for personalized recommendations. Next question: **Building a solution in Azure that uses Azure Cognitive Service for Language to process sensitive customer data. Only specific Azure processes should access the language service, with minimal administration effort.** What should be included? This question is about access control and security, not directly AI capabilities. * **IP Sec Rules:** Used for securing established communication channels between two endpoints, not for restricting specific Azure processes. * **Virtual Network Rules:** This is the correct answer. Virtual Network rules allow you to configure access for specific IP addresses or virtual networks, effectively restricting access to certain Azure processes with minimal administrative overhead. * **Azure Application Gateway and Virtual Network Gateway:** While gateways can manage traffic, they typically require more administrative effort compared to simple Virtual Network rules for this specific scenario. Next question: **Planning a solution in compliance with Responsible AI principles.** Which factor should be considered? Microsoft's Responsible AI principles are crucial here. The correct answer is **accountability**. Other options like cost optimization, scalability, and contracts are general project considerations but not specific Responsible AI principles. Microsoft's six guiding principles for Responsible AI are: 1. **Transparency** 2. **Accountability** 3. **Privacy and Security** 4. **Fairness** 5. **Inclusiveness** 6. **Reliability and Safety** It's important to remember these exact terms. Next question: **Crucial steps when creating Azure AI Services.** * **Configure networking settings:** Not always crucial, sometimes default settings suffice. * **Select appropriate Azure region and right pricing tier:** This is the correct answer. Choosing the right region minimizes latency and addresses data residency requirements (due to laws/regulations). The pricing tier directly impacts cost. * **Setting up the data export schedule:** Not a crucial initial step for *creating* the service. Next question: **Performing predictive maintenance on 100 industrial machines, each with 50 sensors generating data every minute for a year (5,000 time series datasets). The goal is to identify unusual values in each time series to predict machinery failures at a large scale.** Which Azure service should be used? The key phrases are "time series data sets" and "identify unusual values." An "anomaly" is a deviation from regular behavior. * **Azure AI Computer Vision:** Analyzes visual content. Not applicable here. * **Cognitive Search:** A search engine. Not applicable. * **Document Intelligence:** Extracts data from documents. Not applicable. * **Azure AI Anomaly Detector:** This is the correct answer. It's specifically designed to monitor and detect anomalies in time series data, with minimal machine learning expertise required. It supports both batch validation and real-time inference. Next question: **How are client applications typically granted access to an Azure AI Service endpoint?** * **Application must specify a valid subscription key:** This is the correct answer. Access to Azure AI services is primarily based on providing a valid subscription key for authentication. * **Access granted to anonymous users by default:** Incorrect. * **User must enter a username and password associated with the Azure subscription:** While identity is involved, the programmatic access model for client applications typically uses subscription keys. Next question: **In which format are messages exchanged between a client app and an Azure AI Services resource when using a REST API?** The standard format for message exchange in REST APIs for Azure AI Services is **JSON**. Final question: **Using the Face service in Azure to identify gender and age of individuals in a dataset. A Face client object is created in C# using the Face client library. The task is to find rectangular face coordinates and location.** Which method should be called? * **Group Facing:** Groups faces in an image. Not for coordinates. * **Find Similar Async:** Finds similar faces in a list. Not for coordinates. * **Detect With Url Async:** This is the correct answer. This method detects faces in an image (specified by a URL) and returns attributes like face location (coordinates), landmarks, gender, and age. * **Identity Async:** Identifies faces from a face list. Not for coordinates. The `Detect With Url Async` method is specifically designed for detecting faces and retrieving their spatial attributes, along with other characteristics, making it suitable for this task.