A growing library of practical learning paths across foundational, applied, and generative AI.
Learning material for exploration and study—not a single installable package or unified production system.
Collections · Start here · Repository layout · Reproducibility
AI Cookbooks brings two complementary learning collections into one repository while keeping their original folder boundaries and Git history intact:
- Machine learning cookbooks cover mathematics, classical algorithms, neural-network fundamentals, data preparation, data engineering, reinforcement learning, and scikit-learn.
- Generative AI cookbooks cover federated learning and the data, model, training, and evaluation stages of LLM pre-training.
Each notebook or exercise should be treated as an independent learning unit. Review its code, dataset requirements, and runtime assumptions before running it.
Topics and implementations can grow over time; the repository is organized to keep each learning path discoverable without presenting the collection as one fixed curriculum or runtime.
Future tracks can cover additional AI concepts, implementation patterns, and applied learning material. A topic is added only when its notebooks or source material are committed and its place in the catalog is documented; the collections below describe what is available today.
| Collection | Covers | Start here |
|---|---|---|
| Machine Learning Cookbooks | Linear algebra, calculus, PCA, regression, trees, KNN, Naive Bayes, perceptrons, backpropagation, Q-learning, sentiment analysis, preprocessing, relational data workflows, and scikit-learn. | Collection guide |
| Generative AI Cookbooks | Federated learning with Flower plus six progressive notebooks on LLM pre-training. | Collection guide |
| Goal | Recommended route |
|---|---|
| Build ML foundations | Start with mathematics for machine learning, then work through preprocessing, regression, trees, KNN, and Naive Bayes. |
| Understand neural-network mechanics | Explore the perceptron and backpropagation notebooks, then the multivariate-calculus exercises. |
| Work with data systems | Use the relational query playbooks under data engineering. |
| Learn distributed ML | Follow the five federated-learning notebooks in order. |
| Learn the LLM pre-training lifecycle | Follow pre-training-llm from data preparation through evaluation. |
Clone the repository and launch Jupyter from the collection you want to explore:
git clone https://github.com/jayanth-mkv/ai-cookbooks.git
cd ai-cookbooks
python -m venv .venv
# Windows PowerShell: .\.venv\Scripts\Activate.ps1
# macOS/Linux: source .venv/bin/activate
python -m pip install jupyterlab
jupyter labInstall additional libraries only after inspecting the selected notebook. Some exercises require packages such as NumPy, pandas, scikit-learn, Flower, or a local PostgreSQL service; the repository does not claim a tested global dependency set.
ai-cookbooks/
├── full-machine-learning-cookbooks/ # classical ML, mathematics, data, and RL
└── generative-ai-cookbooks/ # federated learning and LLM pre-training
The folder names intentionally preserve the source collection identities. They are organizational boundaries, not separately deployed applications.
| Area | Status |
|---|---|
| Notebook source | Available in both collections. |
| Shared root environment | Not provided; dependencies differ by notebook. |
| Dataset and service access | Required by selected notebooks; inspect code before execution. |
| Automated notebook execution | Not configured. |
| Production readiness | Out of scope; this is a learning repository. |
- Run notebooks in an isolated environment.
- Do not commit local credentials, datasets with restricted terms, or generated training artifacts.
- Check source licenses and course acknowledgements inside individual tracks before reusing material.
- Treat historical notebook outputs as examples, not proof of compatibility with current libraries.