Azure AI-900: Azure AI Fundamentals - Comprehensive Fact Sheet¶

Table of Contents¶

1. Exam Overview
2. AI Workloads and Considerations
3. Machine Learning Principles
4. Computer Vision Workloads
5. Natural Language Processing (NLP)
6. Conversational AI
7. Responsible AI
8. Azure AI Services
9. Study Resources

Exam Overview¶

About the AI-900 Certification¶

The 📖 Azure AI Fundamentals certification - Official certification page with exam details and registration information.

📖 Exam AI-900 Skills Outline - Complete study guide with all measured skills and exam objectives.

📖 Microsoft Learn AI-900 Learning Path - Official training course covering all exam domains.

Exam Format¶

• Questions: 40-60 questions
• Duration: 60 minutes
• Passing Score: 700 out of 1000
• Question Types: Multiple choice, drag-and-drop, hot area, build list
• Languages: Available in multiple languages including English, Japanese, Chinese, Korean, German, French, Spanish, and more

📖 Microsoft Certification Exam Policies - Important policies regarding exam registration, retakes, and accommodations.

AI Workloads and Considerations¶

Understanding Artificial Intelligence¶

📖 What is Artificial Intelligence? - Foundational overview of AI concepts and applications.

📖 Introduction to AI - Microsoft Learn module introducing AI fundamentals.

Common AI Workloads¶

Machine Learning¶

📖 What is Machine Learning? - Overview of machine learning concepts and Azure ML capabilities.

📖 Machine Learning Basics - Fundamentals of ML including supervised and unsupervised learning.

Anomaly Detection¶

📖 Azure AI Anomaly Detector - Service for identifying anomalies in time-series data.

Computer Vision¶

📖 Computer Vision Concepts - Understanding image and video analysis capabilities.

Natural Language Processing¶

📖 Natural Language Processing Overview - Introduction to text analysis and language understanding.

Knowledge Mining¶

📖 Azure AI Search (formerly Cognitive Search) - Extract insights from large volumes of data using AI enrichment.

AI Principles and Considerations¶

📖 Responsible AI Principles - Microsoft's framework for responsible AI development.

📖 AI Ethics and Governance - Understanding ethical considerations in AI systems.

Machine Learning Principles¶

Core Machine Learning Concepts¶

📖 Introduction to Machine Learning - Comprehensive introduction to ML fundamentals.

📖 Types of Machine Learning - Understanding supervised, unsupervised, and reinforcement learning.

Supervised Learning¶

📖 Supervised Learning Concepts - Training models with labeled data for predictions.

Regression¶

• Predicting continuous numerical values
• Examples: Price prediction, demand forecasting
• 📖 Regression in Azure ML - Creating and evaluating regression models.

Classification¶

• Predicting categorical outcomes (binary or multiclass)
• Examples: Spam detection, image classification
• 📖 Classification Models - Building classification solutions.

Unsupervised Learning¶

📖 Unsupervised Learning Overview - Learning patterns from unlabeled data.

Clustering¶

• Grouping similar data points together
• Examples: Customer segmentation, pattern discovery
• 📖 Clustering Algorithms - K-means and hierarchical clustering techniques.

Azure Machine Learning Service¶

📖 Azure Machine Learning Overview - Comprehensive platform for building, training, and deploying ML models.

📖 Azure ML Studio - Web-based interface for ML workflows.

📖 Automated Machine Learning (AutoML) - Automatically selecting algorithms and hyperparameters for optimal models.

📖 Azure ML Designer - Drag-and-drop interface for creating ML pipelines without code.

Model Training and Evaluation¶

📖 Training Models in Azure ML - Best practices for training machine learning models.

📖 Model Evaluation Metrics - Understanding accuracy, precision, recall, F1-score, and RMSE.

📖 Feature Engineering - Selecting and transforming features for better model performance.

Model Deployment¶

📖 Deploying ML Models - Publishing models as web services for real-time predictions.

📖 Real-time vs Batch Inference - Understanding different deployment scenarios.

Computer Vision Workloads¶

Azure AI Vision Service¶

📖 Azure AI Vision Overview - Comprehensive image and video analysis service.

📖 Computer Vision Features - Complete list of available vision capabilities.

Image Analysis¶

📖 Image Analysis API - Analyzing images for content, objects, and features.

📖 Tagging Images - Automatically generating descriptive tags for images.

📖 Object Detection - Identifying and locating objects within images with bounding boxes.

📖 Image Categorization - Classifying images into predefined categories.

📖 Brand Detection - Detecting commercial brands in images and videos.

Optical Character Recognition (OCR)¶

📖 OCR Technology Overview - Extracting text from images and documents.

📖 Read API - Advanced OCR for printed and handwritten text extraction.

Face Detection and Analysis¶

📖 Azure AI Face Service - Detecting and analyzing human faces in images.

📖 Face Detection Concepts - Understanding face detection, attributes, and landmarks.

📖 Facial Recognition - Face verification and identification capabilities.

📖 Limited Access Features - Understanding access restrictions for facial recognition features.

Custom Vision¶

📖 Azure Custom Vision Service - Training custom image classification and object detection models.

📖 Custom Vision Quickstart - Building your first custom classifier.

📖 Custom Object Detection - Creating custom object detection models.

Azure AI Document Intelligence (Form Recognizer)¶

📖 Document Intelligence Overview - Extracting information from documents using AI.

📖 Prebuilt Models - Ready-to-use models for invoices, receipts, ID cards, and business cards.

📖 Custom Document Models - Training models for specific document types.

📖 Layout Analysis - Extracting text, tables, and structure from documents.

Video Analysis¶

📖 Video Analysis Capabilities - Analyzing video content for insights and metadata.

Natural Language Processing (NLP)¶

Azure AI Language Service¶

📖 Azure AI Language Overview - Comprehensive NLP service for text analysis.

📖 Language Service Features - Complete guide to available language capabilities.

Text Analysis¶

📖 Sentiment Analysis - Determining positive, negative, or neutral sentiment in text.

📖 Key Phrase Extraction - Identifying main concepts and topics in text.

📖 Entity Recognition (NER) - Detecting and categorizing entities like people, locations, organizations.

📖 Language Detection - Identifying the language of input text.

📖 Personally Identifiable Information (PII) Detection - Identifying and redacting sensitive information.

Question Answering¶

📖 Question Answering Service - Creating knowledge bases for Q&A scenarios.

📖 Custom Question Answering - Building and deploying custom Q&A solutions.

Conversational Language Understanding (CLU)¶

📖 Conversational Language Understanding - Building natural language understanding models for apps.

📖 Intents and Entities - Core concepts for understanding user input.

Language Translation¶

📖 Azure AI Translator - Real-time text translation across 100+ languages.

📖 Translation Features - Supported languages and translation capabilities.

📖 Custom Translator - Building customized translation models for specific domains.

Speech Services¶

📖 Azure AI Speech Overview - Comprehensive speech-to-text and text-to-speech capabilities.

📖 Speech-to-Text - Converting spoken audio to text transcription.

📖 Text-to-Speech - Generating natural-sounding synthetic speech.

📖 Speech Translation - Real-time translation of spoken language.

📖 Custom Neural Voice - Creating unique brand voices with neural text-to-speech.

📖 Speaker Recognition - Identifying and verifying speakers by voice characteristics.

Conversational AI¶

Azure AI Bot Service¶

📖 Azure Bot Service Overview - Building intelligent conversational bots.

📖 Bot Framework - SDK and tools for bot development.

📖 Bot Framework Composer - Visual authoring canvas for building conversational experiences.

Bot Channels¶

📖 Bot Channels - Connecting bots to Microsoft Teams, Slack, Facebook Messenger, and more.

📖 Direct Line Channel - Custom channel for embedding bots in applications.

Conversational Design¶

📖 Bot Design Principles - Best practices for creating effective conversational experiences.

📖 Dialogs and Prompts - Managing conversation flow and user interactions.

Responsible AI¶

Microsoft's Responsible AI Principles¶

📖 Responsible AI Overview - Microsoft's commitment to responsible AI development.

📖 Microsoft Responsible AI Standard - Core principles and governance framework.

Six Principles of Responsible AI¶

Fairness¶

📖 Fairness in AI Systems - Ensuring AI systems treat all people fairly without bias.

📖 Fairlearn - Toolkit for assessing and improving fairness in ML models.

Reliability and Safety¶

📖 Reliability and Safety - Building AI systems that operate reliably and safely.

Privacy and Security¶

📖 Privacy in AI - Protecting user data and maintaining confidentiality.

📖 Differential Privacy - Techniques for preserving privacy in machine learning.

Inclusiveness¶

📖 Inclusive Design - Creating AI that benefits everyone and engages people of all abilities.

Transparency¶

📖 Transparency in AI - Making AI systems understandable and interpretable.

📖 Model Interpretability - Tools and techniques for explaining model predictions.

Accountability¶

📖 Accountability in AI - Ensuring people are accountable for AI systems and their outcomes.

Responsible AI Tools¶

📖 Responsible AI Dashboard - Unified interface for assessing and debugging AI systems.

📖 Error Analysis - Identifying and diagnosing model errors and weaknesses.

Azure AI Services¶

Azure AI Services Overview¶

📖 Azure AI Services - Suite of AI APIs and services for developers.

📖 AI Services Pricing - Cost information for various AI services.

📖 Creating AI Services Resources - Setting up multi-service or single-service resources.

Authentication and Security¶

📖 Authenticating with AI Services - Using API keys and Azure Active Directory for secure access.

📖 Virtual Networks for AI Services - Network security and isolation for AI resources.

Containers for AI Services¶

📖 AI Services Containers - Running AI services in containerized environments for on-premises or edge scenarios.

Monitoring and Logging¶

📖 Monitoring AI Services - Using Azure Monitor for tracking usage and performance.

📖 Diagnostic Logging - Enabling logs for troubleshooting and analytics.

Study Resources¶

Official Microsoft Resources¶

📖 Microsoft Learn AI-900 Path - Complete learning path with hands-on modules.

📖 AI-900 Practice Assessment - Official practice test to assess readiness.

📖 Microsoft Exam Sandbox - Interactive demo of exam interface and question types.

Documentation and Guides¶

📖 Azure AI Documentation - Central hub for all Azure AI services documentation.

📖 Azure Architecture Center - Reference architectures and best practices for AI solutions.

Hands-On Labs¶

📖 Microsoft Learn Sandbox Exercises - Free interactive labs in browser environment.

📖 Azure Free Account - Get $200 credit and free services for hands-on practice.

Additional Learning¶

📖 AI School - Video tutorials and courses on AI technologies.

📖 Azure AI Blog - Latest updates and insights on Azure AI services.

📖 GitHub Azure AI Examples - Code samples and starter projects for Azure AI services.

Key Concepts Summary¶

Machine Learning Fundamentals¶

• Supervised Learning: Training with labeled data (regression and classification)
• Unsupervised Learning: Finding patterns in unlabeled data (clustering)
• Features and Labels: Input variables (features) used to predict outcomes (labels)
• Training and Testing: Splitting data to train models and validate performance
• Model Evaluation: Using metrics like accuracy, precision, recall, RMSE

Computer Vision Capabilities¶

• Image Classification: Categorizing entire images
• Object Detection: Finding and locating objects with bounding boxes
• OCR: Extracting text from images
• Face Detection: Identifying faces and attributes
• Custom Models: Training specialized vision models

Natural Language Processing¶

• Text Analytics: Sentiment, key phrases, entities, language detection
• Translation: Converting text between languages
• Speech: Speech-to-text, text-to-speech, speech translation
• Conversational AI: Understanding intents and entities
• Question Answering: Building knowledge-base chatbots

Azure AI Services Architecture¶

• Multi-Service Resource: Single endpoint for multiple AI services
• Single-Service Resource: Dedicated resource for specific service
• API Keys: Subscription-based authentication
• Endpoints: Regional service URLs for API calls
• Pricing Tiers: Free and paid tiers with different quotas

Responsible AI Implementation¶

• Bias Detection: Identifying unfair treatment of groups
• Model Transparency: Understanding how models make decisions
• Privacy Protection: Safeguarding user data
• Human Oversight: Maintaining human control and accountability
• Impact Assessment: Evaluating societal implications

Exam Tips¶

Preparation Strategy¶

1. Complete all Microsoft Learn modules in the AI-900 learning path
2. Take practice assessments to identify knowledge gaps
3. Get hands-on experience with Azure AI services using free tier
4. Review documentation for each service's capabilities and use cases
5. Understand when to use each AI service for different scenarios
6. Focus on fundamental concepts rather than deep technical implementation

During the Exam¶

• Read questions carefully and identify key requirements
• Eliminate obviously wrong answers first
• Look for keywords that indicate specific services or capabilities
• Remember that some questions may have multiple correct answers
• Flag difficult questions and return to them later
• Manage your time - don't spend too long on single questions

Common Pitfall Areas¶

• Confusing supervised vs unsupervised learning
• Mixing up different Computer Vision services and their capabilities
• Not understanding the difference between classification and regression
• Forgetting the six principles of Responsible AI
• Confusing Language service features with Speech service features
• Not knowing when to use custom models vs prebuilt models

Quick Reference¶

Service Selector Guide¶

Use Computer Vision when you need to: - Analyze image content (objects, tags, descriptions) - Extract text from images (OCR) - Detect faces and attributes - Moderate content

Use Custom Vision when you need to: - Train custom image classification models - Build custom object detection models - Work with domain-specific images

Use Document Intelligence when you need to: - Extract data from forms and documents - Process invoices, receipts, IDs, business cards - Extract tables and layout information

Use Language Service when you need to: - Analyze text sentiment - Extract key phrases and entities - Detect language - Build Q&A systems - Understand conversational input

Use Translator when you need to: - Translate text between languages - Build multilingual applications - Custom domain translation

Use Speech Service when you need to: - Convert speech to text - Generate speech from text - Translate spoken language - Recognize speakers

Use Bot Service when you need to: - Build conversational interfaces - Create chatbots for multiple channels - Implement complex dialog flows

Use Azure Machine Learning when you need to: - Train custom ML models from scratch - Use AutoML for automated model selection - Deploy models as web services - Manage the complete ML lifecycle

Glossary of Terms¶

• Algorithm: A set of rules or instructions used to solve a problem or complete a task
• API: Application Programming Interface for accessing services programmatically
• Bounding Box: Rectangle coordinates marking object location in images
• Classification: Predicting categorical labels or classes
• Clustering: Grouping similar data points without predefined categories
• Confidence Score: Probability indicating model's certainty in prediction
• Endpoint: URL where API service is accessed
• Entity: Specific item or category identified in text (person, place, organization)
• Feature: Input variable used in machine learning model
• Intent: User's goal or purpose in conversational AI
• Label: Target output or answer in supervised learning
• Model: Trained algorithm that makes predictions on new data
• OCR: Optical Character Recognition for text extraction
• Regression: Predicting continuous numerical values
• SDK: Software Development Kit with libraries and tools
• Sentiment: Emotional tone expressed in text (positive, negative, neutral)
• Training: Process of teaching ML model using data
• Utterance: Example phrase showing how users might express intent

This fact sheet covers all essential topics for the Azure AI-900 certification exam. Focus on understanding core concepts, service capabilities, and appropriate use cases rather than memorizing syntax or implementation details. Good luck with your certification journey!

Last Updated: October 2025