Zero to AI - Free AI/ML Course: Python, LLMs, RAG, Agents & Deep Learning

Zero to AI is a free, open-source AI course and machine learning curriculum for learning Python, data science, deep learning, large language models (LLMs), retrieval-augmented generation (RAG), AI agents, prompt engineering, fine-tuning, and MLOps through 950+ hands-on Jupyter notebooks.

Use the live site for the guided learning experience at zero-to-ai.dev, and use this GitHub repo as the source curriculum, notebook library, and contribution hub.

Quick links: Website • Course Setup • Glossary • GitHub Repo

Click any badge above to start coding in seconds!

From Zero to AI Mastery is still evolving. The core path is already usable, and feedback is welcome as later phases continue to grow.

This repository is best used as a guided curriculum, not as a directory to browse at random. Some phases are mature and portfolio-ready today, while a few later phases are still being filled in.

The canonical reading experience is the live site at zero-to-ai.dev. The GitHub repo is the source of the curriculum (notebooks + MDX).

Zero install path: every notebook can run directly in your browser on the live site via Pyodide — no Python install, no credit card, works on a school laptop or a phone. Just open a phase and run the cells.

If you are new, use this order:

1. Open zero-to-ai.dev/00-course-setup for orientation.
2. Skim zero-to-ai.dev/23-glossary/01_GLOSSARY once before diving in.
3. Pick a path below, then a phase from the site sidebar.

If you are completely new to Python, begin with zero-to-ai.dev/01-python.

Don't try to read all 33 phases. Pick one of these tracks based on your goal:

01-python → 04-token → 11-prompt-engineering → 05-embeddings → 08-rag → 15-ai-agents

01-python → 02-data-science → 03-maths → 06-neural-networks → 09-mlops → 28-practical-data-science → 16-model-evaluation

11-prompt-engineering → 07-vector-databases → 08-rag → 15-ai-agents → 14-local-llms → 12-llm-finetuning → 09-mlops

01-python (run cells in browser on the live site, no install). When you finish Python, come back and pick one of the tracks above.

Run everything on zero-to-ai.dev directly. No GPU, no install, no Anaconda download. The browser does the work. For heavier notebooks (fine-tuning, large models), open them in Google Colab (most globally accessible free GPU).

This comprehensive AI/ML curriculum uses progressive numbered modules covering everything from Python fundamentals to cutting-edge AI systems and advanced research topics. Each module includes hands-on notebooks, projects, and practical applications.

Important context for learners:

• The live site at zero-to-ai.dev is the canonical guided reading experience.
• Source content lives in two parallel trees: next-docs/src/app/<phase>/ (MDX, what the site renders) and jupyter-notebooks/<phase>/ (executable notebooks).
• Not every late-stage phase is equally complete yet. Phases such as 30-inference-optimization/ are still actively being built out.
• next-docs/.next/, next-docs/out/, and other generated folders are build artifacts, not the source curriculum.

Don't forget to:

• ⭐ Star this repo to find it easily later
• 🍴 Fork this repo to track your personal progress

For most learners, the fastest low-friction path is:

1. Install dependencies (see options below).
2. Open the site at zero-to-ai.dev and pick a phase from the sidebar.
3. Open the matching notebook under jupyter-notebooks/<phase>/ and work through it.

• Basic Python knowledge (or start with zero-to-ai.dev/01-python).
• High school maths is enough for most of the course. Phases 3, 24, 25, and 27 go deeper if you want.
• Compute — pick whichever applies:
  • No laptop or low-spec laptop? Run notebooks in the browser on zero-to-ai.dev. Works on Chromebooks and phones. No GPU. No install.
  • 8 GB RAM, any OS? Fine for Phases 1-11 locally. Use Colab/Kaggle for the heavier later phases.
  • 16 GB+ RAM, any OS? Comfortable for most of the curriculum locally. Windows, macOS, and Linux all work — pyproject.toml documents the platform-specific install notes.
  • GPU? Optional. Needed only for local fine-tuning (Phases 12, 14, 24) and large inference (Phase 30). Free cloud GPUs cover those.
• Time: 3-5 hours/week if you're learning casually, 10-15 hours/week if you're studying for a job.

# Clone the repository
git clone https://github.com/PavanMudigonda/zero-to-ai.git
cd zero-to-ai

# Install dependencies with UV (fastest!)
./install_dependencies.sh

# Start learning
jupyter notebook

Optional developer tooling for Phase 31:

# Install Node-based coding tools such as OpenCode
npm install

# Install the dedicated AI developer tools environment
# (used for OpenHands because it currently needs Python 3.12)
INSTALL_AI_DEV_TOOLS=1 ./install_dependencies.sh

# Clone the repository
git clone https://github.com/PavanMudigonda/zero-to-ai.git
cd zero-to-ai

# Create conda environment
conda env create -f environment.yml
conda activate aiml-learning

# Start learning
jupyter notebook

Listed by global accessibility — the higher entries are easiest to use no matter where you are.

• Run on the live site (most accessible — no account, no credit card, no install, works on phones): Open any phase on zero-to-ai.dev and run the cells in-browser via Pyodide. No GPU; not every Python package has a Pyodide wheel. Best for early phases (Python, data science, prompt engineering) and any notebook that doesn't need PyTorch on a GPU.

• Google Colab (free, works in most countries, no credit card): Click - then run this in the first cell:

  !pip install -q -r https://raw.githubusercontent.com/PavanMudigonda/zero-to-ai/main/colab_requirements.txt

  Free GPU is available but rate-limited.

• Kaggle (free GPU, requires phone verification): Click and create a New Notebook, enable Internet in notebook settings, then run:

  !git clone https://github.com/PavanMudigonda/zero-to-ai.git
  %cd zero-to-ai
  !pip install -q -r https://raw.githubusercontent.com/PavanMudigonda/zero-to-ai/main/kaggle_requirements.txt

  Phone verification is restricted in some regions — try Colab first if Kaggle won't verify your number.

• GitHub Codespaces (free monthly hours, requires GitHub account): Click . All dependencies install via install_dependencies.sh. No GPU on the free plan.

• Replit (free tier, browser-based): Click . Packages auto-install on first run via requirements-replit.txt.

• SageMaker Studio Lab (free, US-friendly verification): Click , sign up at studiolab.sagemaker.aws, then run:

  !pip install -q -r requirements.txt

  Account approval can take days and is harder outside the US/EU — use Colab first if you need to start today.

• github.dev (browser editor, no execution): Click . Good for reading and writing code, not running it.

The curriculum lives in two parallel trees, one numbered phase per topic:

zero-to-ai/
├── next-docs/                          # Next.js (Nextra) docs site → published to zero-to-ai.dev
│   ├── src/app/<phase>/page.mdx        # MDX pages, one per phase + per lesson
│   ├── next.config.mjs
│   └── package.json
├── jupyter-notebooks/                  # Executable curriculum content
│   ├── 00-course-setup/                ⭐ Setup & orientation
│   ├── 01-python/                       Python fundamentals
│   ├── 02-data-science/                 NumPy, Pandas, Scikit-learn (278+ notebooks)
│   ├── 03-maths/                        Linear algebra, calculus, statistics (40+ notebooks)
│   │   ├── foundational/
│   │   ├── mml-book/
│   │   ├── islp-book/
│   │   ├── cs229-course/
│   │   └── advanced/
│   ├── 04-token/                        Tokenization (tiktoken, sentencepiece, HF)
│   ├── 05-embeddings/                   Text, multimodal, retrieval embeddings
│   ├── 06-neural-networks/              Deep learning from scratch to Transformers
│   ├── 07-vector-databases/             Chroma, Qdrant, Weaviate, Milvus, pgvector
│   ├── 08-rag/                          Retrieval-Augmented Generation
│   ├── 09-mlops/                        Deployment, monitoring, optimization
│   ├── 10-specializations/              AI Agents, Computer Vision, NLP
│   ├── 11-prompt-engineering/           Advanced prompting, context engineering
│   ├── 12-llm-finetuning/               LoRA, QLoRA, PEFT fine-tuning
│   ├── 13-multimodal/                   Vision, audio, video, realtime multimodal AI
│   ├── 14-local-llms/                   Ollama, llama.cpp, MLX, local serving
│   ├── 15-ai-agents/                    Function calling, MCP, Agents SDK, LangGraph
│   ├── 16-model-evaluation/             Metrics, fairness, LLM-as-judge
│   ├── 17-debugging-troubleshooting/    Profiling, data issues, debugging
│   ├── 18-low-code-ai-tools/            Gradio, Streamlit, Flowise, Langflow, AutoML
│   ├── 19-ai-safety-redteaming/         Security, bias, red teaming
│   ├── 20-real-time-streaming/          Streaming AI, WebSockets, WebRTC
│   ├── 21-quizzes/                      Self-check questions
│   ├── 22-references/                   Papers, repos, external materials
│   ├── 23-glossary/                     AI/ML terminology & concepts
│   ├── 24-advanced-deep-learning/       GANs, VAEs, NeRF, Diffusion, BNNs
│   ├── 25-reinforcement-learning/       MDP, Q-Learning, Policy Gradients, Actor-Critic
│   ├── 26-time-series-analysis/         ARIMA, Prophet, LSTM, Transformer forecasting
│   ├── 27-causal-inference/             DAGs, Experimental Design, Observational Methods
│   ├── 28-practical-data-science/       Interview prep & hands-on practice
│   ├── 29-ai-hardware-llm-validation/   Silicon validation for AMD, NVIDIA, Qualcomm, TPU, Apple Silicon
│   ├── 30-inference-optimization/       KV cache, vLLM, TensorRT-LLM, quantization
│   ├── 31-ai-powered-dev-tools/         VS Code AI setup, MCP deep dive, tool workflows
│   ├── 32-cheatsheets/                  Quick references across tooling/cloud/AI/ML
│   └── 33-roadmaps/                     Suggested learning roadmaps
├── scripts/                            # Utility scripts (notebook sync, MCP server, scrapers)
├── pyproject.toml                      # Primary Python deps (uv-managed)
├── requirements*.txt                   # Per-target deps (Colab, Kaggle, Replit, AI dev tools)
├── environment.yml                     # Conda alternative
└── install_dependencies.sh             # One-shot installer (UV-based)

Each phase under next-docs/src/app/ mirrors the same phase under jupyter-notebooks/. The site renders MDX; the notebooks are runnable.

• Mature core path: 02-data-science/ through 19-ai-safety-redteaming/
• Strong advanced depth: 24-advanced-deep-learning/, 28-practical-data-science/
• Still being expanded: 20-real-time-streaming/, 21-quizzes/, 22-references/, 30-inference-optimization/
• Build artifacts (not source): next-docs/.next/, next-docs/out/, node_modules/, .venv/

Once you have oriented yourself:

1. Open the live site at zero-to-ai.dev and pick a phase from the sidebar.
2. Open the matching notebook under jupyter-notebooks/<phase>/ to run it locally.
3. Pair the two: read the explanation on the site, execute the notebook beside it.

We welcome contributions from the community! Here's how you can help:

• 🐛 Report Issues: Found a bug or error? Open an issue
• ✨ Suggest Features: Have an idea? Share it in Discussions
• 📝 Improve Documentation: Fix typos, clarify explanations
• 💻 Add Examples: Contribute new notebooks or examples
• 🌍 Translations: Help translate content to other languages

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

See CONTRIBUTING.md for the full guide, including how to keep jupyter-notebooks/ and next-docs/ in sync, and a note on translations. For larger changes, please open an issue first to discuss the approach.

This project follows a Code of Conduct. By participating you agree to uphold it.

This learning path is inspired by and integrates content from:

• Microsoft Learn - Official AI/ML curriculum
• Microsoft for Beginners Series - AI, ML, GenAI, Agents courses
• Stanford University - CS229, CS224N, CS231N courses
• 3Blue1Brown - Visual mathematics explanations
• Andrej Karpathy - Neural networks from scratch
• DeepLearning.AI - Practical AI courses
• Fast.ai - Practical deep learning approach

Special Thanks to all contributors, educators, and the open-source community!

This project is licensed under the MIT License.

🚀 From Zero to AI - Your Journey Starts Here!

"The best time to start learning AI was yesterday. The second best time is now."

Questions? Open an issue or start a discussion!

Found this helpful? ⭐ Star this repo and share it with others!