Microsoft

Generative AI for Beginners

Build generative AI apps across 21 lessons with Python and TypeScript samples, for developers new to the field.

Taught by Microsoft Cloud Advocates at MicrosoftFreeVerified 11 Oct 2026

Open the course

From Course Setup to Building with Meta Models: 22 lessons
OpenAI's GPT API for apps, agents and voice: OpenAI Platform
Before you startBasic Python or TypeScript helps; a GitHub account and access to a model service (Azure OpenAI, Microsoft Foundry, OpenAI API or Foundry Local).: Beginner level
21 lessons: Self-paced

Generative AI for BeginnersFreeOpen the course

What you’ll learn

Learn what generative AI and LLMs are, and how to pick the right model for a task.
Write better prompts, from the basics to advanced techniques.
Build text generation, chat, search, image generation and low-code AI apps.
Add function calling, RAG with vector databases, open source models and AI agents to an app.
Plan for real use: responsible AI, UX design, security, the app lifecycle and fine-tuning.

Skills you’ll gain

PromptingPythonAgentsFine-tuningLLMsResponsible AIBuild With an API

Tools you'll learn

OpenAI PlatformOpenAI's GPT API for apps, agents and voicePaid

Details to know

No certificateMicrosoft lists none
Free to takeNo fee on github.com
Taught in English55 languages available
VerifiedOct 2026

Lessons

21 lessons

Generative AI for Beginners is an open-source course from Microsoft Cloud Advocates, published on GitHub and now in its third version. It has 21 lessons after a setup lesson, and each one stands alone, so you can start with any topic.

Learn lessons explain a concept, and Build lessons add code in Python and TypeScript for Azure OpenAI, Microsoft Foundry Models or the OpenAI API. Every lesson pairs a written guide with a short video, though some videos are still to come, and links to further reading.

Across 21 lessons and a setup lesson, you will:

Course Setup: Fork the repo, open it in a codespace or on your machine, and connect a model service.
Lessons 01 to 03: Learn what LLMs are, how to compare and choose one, and how to use them responsibly.
Lessons 04 and 05: Write prompts from the basics up to few-shot, chain-of-thought and self-refine.
Lessons 06 to 09: Build text generation, chat, embeddings search and image generation apps.
Lessons 10 and 11: Build low-code apps in Power Platform and connect models to your code with function calling.
Lessons 12 to 14: Design the user experience, secure the app and manage it through its lifecycle.
Lessons 15 to 18: Ground answers with RAG and vector databases, use open models, build agents and fine-tune.
Lessons 19 to 21: Work with small language models, Mistral models and Meta's Llama models.

Applied learning project

Nine Build lessons come with code samples, and many lessons end in an assignment, often set at the course's education startup. Several lessons include reference solutions to check your work against.

Recipe generator: Build a text app that suggests recipes, then add filters and a shopping list.
Video search: Search transcripts of Microsoft's AI Show videos with embeddings and jump to the answer.
Monument images: Generate images of landmarks in new settings, with metaprompts as guardrails.
Low-code apps: Build a student assignment tracker in Power Apps and an invoice processing flow in Power Automate.
Course finder: Extend a chatbot's function calls so it finds courses for learners and handles errors.
Agent meeting: Simulate a startup meeting where department agents question your product pitch.

By the end, you can build small generative AI apps that use prompts, embeddings, RAG, function calling and agents, with responsible AI and security in mind.

Course SetupLesson 1Lesson details

What you’ll learn

Fork the repo and open it in GitHub Codespaces, or run it on your own machine.
Set up keys for Azure OpenAI or the OpenAI API in a .env file.

Introduction to Generative AI and LLMsLesson 2Lesson details

What you’ll learn

Trace how AI grew from statistical methods and neural networks into today's generative models.
Learn how large language models produce text and where they fit in education.
Sketch your own AI startup idea in a 300-word write-up.

Exploring and comparing different LLMsLesson 3Lesson details

What you’ll learn

Compare model types: open and proprietary, embedding, image and text, encoder-decoder and decoder-only.
Test and iterate on models to pick one for your use case.
Choose between prompting with context, RAG, fine-tuning or training your own model.

Using Generative AI ResponsiblyLesson 4Lesson details

What you’ll learn

Spot risks such as hallucinations, harmful content and unfair output.
Measure and reduce harms, and keep a responsible AI solution running well.

Understanding Prompt Engineering FundamentalsLesson 5Lesson details

What you’ll learn

Learn what prompt engineering is and why the same prompt behaves differently across models.
Build prompts from instructions, primary content, examples, cues and templates.
Practice in a Jupyter notebook against an OpenAI endpoint.

Creating Advanced PromptsLesson 6Lesson details

What you’ll learn

Apply zero-shot, few-shot, chain-of-thought, generated knowledge and least-to-most prompting.
Get the model to critique and refine its own answers.
Control how varied the output is with temperature.

Building Text Generation ApplicationsLesson 7Lesson details

What you’ll learn

Build your first text generation app with the openai library.
Turn it into a recipe generator with filters and a shopping list.
Adjust tokens and temperature to change the output.

Building Chat ApplicationsLesson 8Lesson details

What you’ll learn

Plan how to add chat to an existing system with SDKs and APIs.
Customize a chat app for one domain and track metrics for its quality.
Work through exercises from first chat prompts to classifying and summarizing text.

Building Search Apps Vector DatabasesLesson 9Lesson details

What you’ll learn

Tell semantic search from keyword search.
Turn text into embeddings and compare them with cosine similarity.
Build a search app that finds the point in a video that answers a question.

Building Image Generation ApplicationsLesson 10Lesson details

What you’ll learn

Generate images in Python with the gpt-image model family.
Edit images and set limits on what gets made with metaprompts.

Building Low Code AI ApplicationsLesson 11Lesson details

What you’ll learn

Use Copilot in Power Apps and Power Automate to build from plain-language prompts.
Build a student assignment tracker and an AI invoice processing flow.
Create agents in Microsoft Copilot Studio.

Integrating External Applications with Function CallingLesson 12Lesson details

What you’ll learn

Learn what function calling is for and where it helps.
Set up a function call with the Azure OpenAI Service.
Wire function calls into a chatbot that finds courses for learners.

Designing UX for AI ApplicationsLesson 13Lesson details

What you’ll learn

Design for usability, reliability, accessibility and a pleasant experience.
Build trust with explainability and user control, and collect feedback.

Securing Your Generative AI ApplicationsLesson 14Lesson details

What you’ll learn

Learn the main threats and risks to AI systems.
Protect data and test your app with AI red teaming.

The Generative AI Application LifecycleLesson 15Lesson details

What you’ll learn

Learn how LLMOps differs from MLOps.
Follow the lifecycle from ideation through building to running the app.
Track quality, harm, honesty, cost and latency with tools such as PromptFlow.

Retrieval Augmented Generation (RAG) and Vector DatabasesLesson 16Lesson details

What you’ll learn

Ground a model's answers in your own data with retrieval augmented generation.
Store embeddings in a vector database, then retrieve and re-rank results.
Measure how well the app answers with evaluation metrics.

Open Source Models and Hugging FaceLesson 17Lesson details

What you’ll learn

Learn what makes a model open source and what open models offer.
Explore Llama 2, Mistral and Falcon on Hugging Face and the Microsoft Foundry catalog.

AI AgentsLesson 18Lesson details

What you’ll learn

Learn what AI agents are and how they keep state and use tools.
Compare LangChain, AutoGen, Microsoft Agent Framework, TaskWeaver and JARVIS.
Build a meeting simulation where department agents question your product pitch.

Fine-Tuning LLMsLesson 19Lesson details

What you’ll learn

Learn what fine-tuning is and when it pays off over prompting or RAG.
Fine-tune a model on Microsoft Foundry, which uses LoRA under the hood.
Walk through a fine-tuning tutorial of your choice.

Building with SLMsLesson 20Lesson details

What you’ll learn

Learn how small language models differ from large ones and where they fit.
Run Microsoft's Phi-3 and Phi-3.5 models through cloud APIs or locally with ONNX Runtime.

Building with Mistral ModelsLesson 21Lesson details

What you’ll learn

Compare Mistral Large, Mistral Small and Mistral NeMo and what each is for.
Run code samples, including RAG with Mistral Large 2 and a tokenizer comparison.