All certifications / Azure AI Engineer / Lessons
Azure AI Engineer AI-102 lessons
Study Azure AI Engineer for free
A week-by-week plan with every lesson, quizzes, checkpoint tests, a practice exam and hands-on labs.
Open the Azure AI Engineer study planA week-by-week plan with every lesson, quizzes, checkpoint tests, a practice exam and hands-on labs.
Domain 1: Plan and manage an Azure AI solution
- Choosing the right Azure AI service for a workload: Azure OpenAI, AI Search, Vision, Language, Speech, Translator and Document Intelligence
- Azure AI Foundry hubs, projects and resources vs single-service and multi-service Azure AI services resources
- Planning model deployments: regions, model availability, deployment types (Standard, Global, Data Zone, Provisioned) and quotas
- Deploying Azure AI resources with the portal, Azure CLI, ARM templates and Bicep
- Endpoints, keys and SDKs: calling Azure AI services with REST and the Python and C# SDKs
- Authentication: API keys vs Microsoft Entra ID, managed identities and Cognitive Services RBAC roles
- Protecting keys and networks: Azure Key Vault, key rotation, private endpoints and network restrictions
- Running Azure AI services in containers: connected and disconnected containers and billing settings
- Monitoring and cost: Azure Monitor metrics, diagnostic logs, alerts, pricing tiers and budgets
- Responsible AI principles and Azure AI Content Safety: harm categories, severity levels and blocklists
- Content filters, prompt shields and groundedness detection for generative AI applications
Domain 2: Implement generative AI solutions
- Azure OpenAI and the Azure AI Foundry model catalog: chat, reasoning, embedding and image models, and how to choose and deploy one
- Chat completions: system, user and assistant messages, and calling a deployment with the Azure OpenAI SDK
- Tuning model output with parameters: temperature, top_p, max tokens, stop sequences and penalties
- Prompt engineering techniques: clear instructions, few-shot examples, output formats and step-by-step reasoning
- Retrieval augmented generation (RAG): grounding a model in your own data with Azure AI Search
- Embeddings and vector retrieval for RAG: chunking documents, embedding models and similarity search
- Prompt flow in Azure AI Foundry: flows, nodes, connections, variants and deployment
- Evaluating generative AI apps: groundedness, relevance, coherence, fluency and safety evaluations
- Fine-tuning vs prompt engineering vs RAG: when each approach fits and how fine-tuning data is prepared
- Generating images and working with multimodal chat models that accept images
- Operating generative AI apps: token usage, rate limits, latency, tracing and monitoring
Domain 3: Implement an agentic solution
- What an AI agent is: model, instructions and tools, and when an agent fits better than a plain chat app
- Azure AI Foundry Agent Service: agents, threads, messages and runs
- Agent tools: file search, code interpreter, Azure AI Search, OpenAPI and function calling
- Function calling: describing functions with JSON schema and handling tool calls in code
- Building agents in code with Semantic Kernel, AutoGen and the Microsoft Agent Framework
- Multi-agent solutions: orchestration patterns, handoffs and connected agents
- Testing, securing and monitoring agents: tracing, human approval and least-privilege tools
- Deploying agents and integrating them into applications
Domain 4: Implement computer vision solutions
- Azure AI Vision Image Analysis: captions, dense captions, tags, objects, people and smart crops
- Reading printed and handwritten text with the OCR (Read) feature of Azure AI Vision
- Custom Vision: image classification (multiclass vs multilabel) vs object detection
- Training and evaluating a Custom Vision model: tagging images, iterations, precision, recall and mAP
- Publishing and consuming a Custom Vision model: prediction resource, published iterations and exporting compact models
- Azure AI Face: face detection, attributes and Limited Access features
- Analyzing video with Azure AI Video Indexer
- Choosing between prebuilt Image Analysis, Custom Vision and multimodal models for a vision task
Domain 5: Implement natural language processing solutions
- Azure AI Language text analysis: language detection, key phrases, entities, entity linking and sentiment with opinion mining
- Detecting and redacting personally identifiable information (PII) with Azure AI Language
- Translating text and documents with Azure AI Translator and Custom Translator
- Speech to text with Azure AI Speech: real-time, continuous and batch transcription
- Text to speech with neural voices and SSML for pronunciation, pauses, rate and style
- Speech translation, intent recognition and keyword recognition with the Speech SDK
- Conversational language understanding (CLU): intents, entities, utterances, training and deployment
- Custom question answering: projects, sources, multi-turn prompts, synonyms and confidence thresholds
- Custom text classification and custom named entity recognition
- Custom speech models and custom neural voice: when to train them and responsible use limits
Domain 6: Implement knowledge mining and information extraction solutions
- Azure AI Search components: data sources, indexers, indexes, skillsets, service tiers, replicas and partitions
- Designing a search index: key field, field attributes (searchable, filterable, sortable, facetable, retrievable), analyzers and suggesters
- AI enrichment with built-in skills: document cracking, OCR, image analysis, entities and key phrases in a skillset
- Custom skills: calling an Azure Function or web API from a skillset
- Querying an index: simple and full Lucene syntax, OData filters, facets, sorting and paging
- Semantic ranking, vector search and hybrid queries in Azure AI Search
- Knowledge store: projections to tables, objects and files in Azure Storage
- Azure AI Document Intelligence prebuilt models: read, layout, invoice, receipt and ID document
- Custom Document Intelligence models: template vs neural, custom classifiers and composed models
- Azure AI Content Understanding: extracting fields from documents, images, audio and video with analyzers