Want to Master AI? Start with These 15 GitHub Repositories

The Gemini API Cookbook is a collection of practical guides and examples for Google's Gemini API. Explore prompting, embeddings, function calling, multimodal inputs and other API workflows. Run a quickstart first, then adapt it to a small task of your own.

The Gemma Cookbook focuses on Google's Gemma open models. Its examples can help you explore model use, customization and implementation. Check the repository's README for the latest examples and supported workflows, since model tooling can change quickly.

The Generative AI on Google Cloud repository brings together notebooks, sample code and applications. It is useful when you want to understand how generative AI features can be developed and managed within cloud workflows. Some examples may require a cloud project or credentials, so review setup instructions before running them.

The Agent Development Kit for Python is an open-source toolkit for building, evaluating and deploying AI agents. Its documentation and samples can help Python developers learn how to define agents, connect tools and inspect results. Start with a single-agent example before exploring multi-agent workflows.

The ADK Recipes repository contains practical examples built with the Agent Development Kit. Use a recipe to see how an agent project is organized, then follow its components back to the relevant framework documentation. Treat sample code as a learning starting point and confirm its requirements before adapting it for production.

Google DeepMind's educational repository offers learning resources for people exploring AI and machine learning. Use it to strengthen the ideas behind the code you encounter elsewhere, and work through the material at a pace that lets you explain each concept in your own words.

The TensorFlow repository is the home of the TensorFlow framework and a gateway to its learning resources. If you're new to machine learning, begin with a tutorial that includes a complete, runnable example rather than trying to understand the entire framework at once.

TensorFlow Models includes implementations and examples for machine learning tasks such as computer vision and language processing. It is especially useful once you can read basic model code and want to study how a solution is structured. Follow the documentation for each model because setup and dependencies can differ.

JAX provides tools for high-performance numerical computing, including automatic differentiation and transformations that can compile or vectorize functions. Work through an introductory example before using it for a larger model so you understand how its operations differ from a standard NumPy workflow.

Flax provides neural-network libraries and tools within the JAX ecosystem. Its examples can help you see how models are defined and trained. Learn the basic structure of a neural network first, then use a Flax example to connect those concepts to implementation.

Optax is a gradient-processing and optimization library for JAX. Optimizers influence how a model's parameters change during training, so this repository is most useful once you understand the basics of loss functions, gradients and training loops.

The Deep Learning Tuning Playbook explains a systematic approach to improving deep learning experiments. Its practical guidance can help you plan what to change, compare runs and avoid drawing conclusions from poorly controlled experiments. The process is useful across frameworks, not just one codebase.

Vertex AI Samples offers code examples and notebooks for machine learning workflows on Google Cloud. Browse by task and pick one example with clear setup steps. Some notebooks may use billable cloud resources, so check the listed requirements and costs before launching services.

MediaPipe provides customizable machine learning solutions for live and streaming media. It is a strong option if you want to build something interactive with vision, audio or other media inputs. Start with a task example that runs on a device or platform you already use.

Dopamine is a research framework for reinforcement learning. Reinforcement learning studies how an agent can learn decisions through interaction and feedback. This topic has a steeper learning curve, so it helps to understand basic machine learning and reward-based learning before diving into the experiments.