A 21-lesson course on building with generative AI, not just reading about it. Each lesson combines written concepts with code samples and an assignment, moving from model and prompt fundamentals toward retrieval, agents, responsible AI, and application architecture.
- Files
- 11,067
- Markdown
- 2,268
- Size
- 771 MB
- License
- MIT
- Last push
- 2026-09-17
Best for Developers who can code but need a practical map of the generative-AI application stack.
What’s inside - Prompt engineering
- Text & image apps
- RAG & vector search
- AI agents
Why it’s worth checking out The repo is approachable without being a thin product tutorial. It gives newer builders a vocabulary for making architectural choices and includes enough runnable material to turn passive reading into experiments.
Clone it git clone --depth 1 https://github.com/microsoft/generative-ai-for-beginners.git
This 12-week, 26-lesson curriculum focuses on classic machine learning through regional datasets and small projects. Written lessons, pre- and post-lesson quizzes, assignments, visualizations, and code are kept together and translated by the community.
- Files
- 17,444
- Markdown
- 5,581
- Size
- 1.2 GB
- License
- MIT
- Last push
- 2026-09-15
Best for Beginners who want to understand the standard ML workflow before jumping straight to large language models.
What’s inside - Regression
- Classification
- Clustering
- NLP & time series
Why it’s worth checking out Its scope is refreshingly disciplined: framing data, training familiar model families, and evaluating results. The recurring lesson structure makes it easier to finish than an unranked pile of notebooks or external course links.
Clone it git clone --depth 1 https://github.com/microsoft/ML-For-Beginners.git
A broad 12-week introduction to artificial intelligence with 24 lessons, labs, quizzes, and diagrams. It connects neural networks to computer vision and natural-language processing while also giving older and alternative AI techniques useful context.
- Files
- 13,705
- Markdown
- 3,752
- Size
- 2.0 GB
- License
- MIT
- Last push
- 2026-09-16
Best for Students and developers who want a survey of AI before choosing a narrower machine-learning specialty.
What’s inside - Neural networks
- Computer vision
- Natural language processing
- Responsible AI
Why it’s worth checking out The curriculum is useful as a map: it shows how major AI subfields relate, then gives readers executable notebooks to test the ideas. That combination makes the repository more durable than a trend-focused model guide.
Clone it git clone --depth 1 https://github.com/microsoft/AI-For-Beginners.git
DAIR.AI’s guide is a maintained field manual for prompting and context engineering. The repository contains technique explainers, model notes, application patterns, safety topics, papers, and notebooks instead of limiting itself to a gallery of clever prompts.
- Files
- 1,559
- Markdown
- 1,111
- Size
- 80 MB
- License
- MIT
- Last push
- 2026-03-11
Best for Practitioners who need to compare prompting, retrieval, and agent patterns across models and use cases.
What’s inside - Few-shot & chain-of-thought
- RAG
- Agents & tool use
- Risks & evaluation
Why it’s worth checking out It is one of the better bridges between fast-moving research terminology and day-to-day implementation. The guide is most valuable when used critically: follow its paper links, then test the claimed pattern against your own model and task.
Clone it git clone --depth 1 https://github.com/dair-ai/Prompt-Engineering-Guide.git
The complete fastai book is published here as executable Jupyter notebooks. It starts with useful models early, then opens up the machinery underneath: data pipelines, training loops, computer vision, NLP, tabular modeling, recommenders, and deployment.
- Files
- 490
- Markdown
- 33
- Size
- 55 MB
- License
- See repo
- Last push
- 2026-09-13
Best for Comfortable Python programmers who prefer learning deep learning by running and changing real notebooks.
What’s inside - Computer vision
- NLP
- Tabular & recommenders
- Training foundations
Why it’s worth checking out The notebooks are both prose and software, which makes this repository unusually easy to interrogate. Readers can search an explanation, run the adjacent code, and gradually move from high-level APIs toward the underlying math and implementation.
Clone it git clone --depth 1 https://github.com/fastai/fastbook.git
Sebastian Raschka’s companion repository builds a GPT-style language model step by step in PyTorch. Chapter folders combine explanatory notebooks, compact implementations, exercises, and setup notes from tokenization through pretraining and fine-tuning.
- Files
- 358
- Markdown
- 76
- Size
- 7 MB
- License
- See repo
- Last push
- 2026-09-17
Best for Python and PyTorch users who want to understand what happens below an LLM API.
What’s inside - Tokenization & data
- Attention
- GPT architecture
- Pretraining & fine-tuning
Why it’s worth checking out It reduces a large modern model to parts a single reader can run and inspect. The value is not in producing a frontier model; it is in building a reliable mechanical understanding of the training and inference pipeline.
Clone it git clone --depth 1 https://github.com/rasbt/LLMs-from-scratch.git
An open textbook whose chapters are runnable notebooks, covering linear models, convolutional networks, optimization, attention, and transformers. The prose and the code sit in the same file, so every equation has an implementation beside it. Three separate licence files cover the book text and the sample code, which is why GitHub cannot classify it with one SPDX id.
- Files
- 1,027
- Markdown
- 209
- Size
- 144 MB
- License
- See repo
- Last push
- 2024-08-18
Best for Readers who want the mathematics under modern architectures with code they can execute while reading.
What’s inside - Linear models & MLPs
- Convolutional networks
- Optimization
- Attention & transformers
Why it’s worth checking out Most deep-learning theory lives in PDFs you cannot run. This one clones, executes, and stays readable offline. Its last push was August 2024 because the book finished, not because it was abandoned; the derivations it teaches have not changed since.
Clone it git clone --depth 1 https://github.com/d2l-ai/d2l-en.git
The full text of Hugging Face’s course on transformers, datasets, tokenizers, and fine-tuning, kept as Markdown chapters in the repository rather than only on the website. Translations for a dozen languages live alongside the English source.
- Files
- 1,444
- Markdown
- 1,139
- Size
- 20 MB
- License
- Apache-2.0
- Last push
- 2026-09-17
Best for Developers moving from calling an LLM API to fine-tuning and evaluating models on their own data.
What’s inside - Transformer models
- Datasets & tokenizers
- Fine-tuning
- Evaluation
Why it’s worth checking out Its star count is the lowest on this page and its Markdown volume is the highest, because people take the course on the website and never visit the repository. That gap is the argument for measuring these repositories instead of ranking them by popularity.
Clone it git clone --depth 1 https://github.com/huggingface/course.git
Course units on building agents: tool calling, the current frameworks, MCP, and evaluation, with runnable notebooks and a final project. Kept current against the frameworks it teaches rather than frozen at one release.
- Files
- 441
- Markdown
- 421
- Size
- 3 MB
- License
- Apache-2.0
- Last push
- 2026-09-15
Best for People who can prompt a model well and now want it to call tools in a loop without guessing at the failure modes.
What’s inside - Agent fundamentals
- Tool calling & MCP
- Frameworks
- Evaluation
Why it’s worth checking out Agent material dates faster than anything else here, so a repository that keeps pace is worth more than a larger one that stopped. It also explains why agents fail, which most framework documentation skips.
Clone it git clone --depth 1 https://github.com/huggingface/agents-course.git
