All certifications / Azure AI Fundamentals / Lessons
Azure AI Fundamentals AI-900 lessons
Study Azure AI Fundamentals for free
A week-by-week plan with every lesson, quizzes, checkpoint tests, a practice exam and hands-on labs.
Open the Azure AI Fundamentals study planA week-by-week plan with every lesson, quizzes, checkpoint tests, a practice exam and hands-on labs.
Domain 1: Describe AI workloads and considerations
- What AI is: models that learn patterns from data, and how that differs from rule-based software
- Common AI workloads: prediction and forecasting, anomaly detection, computer vision, NLP, document processing and generative AI
- Computer vision workloads: image classification, object detection, optical character recognition and facial analysis
- Natural language processing and speech workloads: text analysis, translation, speech recognition and conversational AI
- Document processing and knowledge mining workloads: extracting fields from forms and making content searchable
- Generative AI workloads: creating text, code and images from prompts, copilots and agents
- Responsible AI principles: fairness, and reliability and safety
- Responsible AI principles: privacy and security, and inclusiveness
- Responsible AI principles: transparency and accountability
- Responsible AI in practice: identifying harms, human oversight, Limited Access features and transparency notes
Domain 2: Fundamental principles of machine learning on Azure
- Features and labels, training and validation data, and how a model is trained and then used for inference
- Regression: predicting a numeric value, with evaluation metrics MAE, RMSE and R²
- Binary and multiclass classification, with accuracy, precision, recall, F1 and the confusion matrix
- Clustering: grouping unlabeled data, and supervised vs unsupervised learning
- Deep learning: neural networks, weights and layers, and why they suit images, speech and language
- The transformer architecture: tokens, embeddings and attention
- Azure Machine Learning: workspace, studio, data assets, compute instances and compute clusters
- Automated machine learning (AutoML) and the Azure Machine Learning designer
- Deploying models to endpoints for real-time or batch inference, and responsible AI tools in Azure Machine Learning
Domain 3: Computer vision workloads on Azure
- Image classification vs object detection vs semantic segmentation
- How computer vision models work: pixels, filters, convolutional neural networks and multimodal models
- Azure AI Vision image analysis: captions, dense captions, tags, object detection, people detection and smart crops
- Optical character recognition (OCR) with the Azure AI Vision Read feature
- Face detection and analysis with Azure AI Face, and the Limited Access policy for identification and verification
- Azure AI Document Intelligence: prebuilt models (invoices, receipts, IDs), the layout model and custom models
- Creating and using Azure AI services resources: multi-service vs single-service resources, endpoints, keys and the free F0 tier
- Choosing the right Azure vision service for a scenario
Domain 4: Natural language processing workloads on Azure
- Language detection, sentiment analysis and opinion mining
- Key phrase extraction, named entity recognition, entity linking and PII detection
- Tokenization, embeddings and semantic similarity: how text becomes numbers
- Summarization and custom text classification in Azure AI Language
- Speech recognition (speech to text) and speech synthesis (text to speech) with Azure AI Speech
- Azure AI Translator for text and documents, and speech translation
- Conversational language understanding: utterances, intents and entities
- Question answering: building a knowledge base from FAQs for a bot
- Choosing the right Azure language or speech service for a scenario
Domain 5: Generative AI workloads on Azure
- How large language models generate text: tokens, next-token prediction and pretraining
- Common generative AI scenarios: copilots, chat assistants, content drafting, summarization, code and image generation
- Prompt engineering: system messages, user prompts, few-shot examples and clear instructions
- Grounding and retrieval augmented generation (RAG) with your own data
- Model settings: temperature, top_p and maximum response length
- Microsoft Foundry (formerly Azure AI Foundry): hubs and projects, the model catalog and the playgrounds
- Azure OpenAI models in Foundry: GPT chat models, embeddings models and image generation models
- AI agents: models with instructions, tools and knowledge, and the Foundry Agent Service
- Responsible generative AI: identify, measure, mitigate and operate; content filters and Azure AI Content Safety
- Evaluating generative AI apps: groundedness, relevance, fluency and safety evaluations, and red teaming