Master AI: Top 10 GitHub Projects from Microsoft & Experts!

By /

Top GitHub LLM Repositories for AI Engineers in 2025

Are you an AI engineer looking for resources to put your skills to the test? With vast amounts of information available, finding the right solutions can be challenging. This article presents a curated list of ten GitHub LLM repositories every AI engineer should know. These are hands-on, real-world projects developed by experts from Microsoft, Karpathy, and various open-source communities.

Whether you’re new to machine learning, deeply involved in large language models, or deploying AI agents, these repositories provide simple code, guided projects, and industry-relevant domains to explore. Consider this your guide to smarter, faster, and better AI development, from learning to building and deploying.

Video link: https://youtube.com/shorts/67AK8bqQG-Q?si=yedSfw6hkjK_w5dn

Table of Contents

Machine Learning for Beginners

Machine Learning for Beginners is a 12-week learning plan created by Microsoft. It teaches the basics of machine learning with real-world data and the scikit-learn library. The course is structured like a classroom, covering supervised and unsupervised learning, classification, regression, clustering, and time series analysis. Each module includes interactive Jupyter notebooks, activities, and quizzes to ensure understanding. This repository simplifies complex machine learning concepts into digestible topics, allowing individuals to learn through practice and experimentation.

Best For:

  • Complete beginners who want a structured way to start learning about machine learning.
  • Educators teaching applied ML.
  • Self-learners who wish to learn from real data and build a portfolio.

GitHub Repository: https://github.com/microsoft/ML-For-Beginners

AI for Beginners

AI for Beginners is an extension of the ML base, designed to introduce students to AI. This course explores deep learning, natural language processing, computer vision models, and transformers. Also created by Microsoft, this 12-week course offers tools like PyTorch and TensorFlow, enabling students to learn foundational AI principles through hands-on practice and interactive labs. It balances algorithmic principles with ethical AI considerations, model deployment, and real-world implementation. It’s best for students transitioning from standard ML to AI.

Best For:

  • Students transitioning from ML to AI.
  • Developers wanting to work with neural networks and transformer models.
  • Students wanting experience and project exposure to modern AI applications.

GitHub LLM Repository: https://github.com/microsoft/AI-For-Beginners

Neural Networks: Zero to Hero

A hands-on dive into the inner workings of deep learning, Neural Networks: Zero to Hero, created by Andrej Karpathy, focuses on building neural networks and GPT-style models from scratch using only Python and NumPy, without high-level libraries. Karpathy breaks down complex concepts like backpropagation, gradient descent, and self-attention into easy-to-learn lessons with code. The highlight is the mini-GPT implementation, which covers how transformers function at a low level.

Best For:

  • Engineers and researchers wanting to learn about deep learning from first principles.
  • People wanting to implement neural networks from scratch.
  • Curious learners who love looking at low-level code.

GitHub Repository: https://github.com/karpathy/nn-zero-to-hero

Deep Learning Paper Implementations

This repository is a curated collection of PyTorch implementations of the latest deep learning papers, including GANs, Transformers, Diffusion Models, and more. Its goal is to assist developers who want to go beyond reading deep learning papers and implement the articles. Each model is implemented clearly and concisely, often achieving the same results as referenced in the paper. With this repository, engineers can reproduce experiments, understand inventions, and extend modern state-of-the-art architectures in the fields of generative AI and computer vision.

Best For:

  • Reproducing state-of-the-art results from leading ML papers.
  • Learning new architectures with actual code.
  • Extending or modifying advanced deep learning models.

GitHub LLM Repository: https://github.com/lucidrains

Made With ML

Made With ML is a complete curriculum created for the entire machine learning lifecycle, from design and development to deployment and monitoring. Built by Goku Mohandas, it focuses on practical skills like data versioning (DVC), continuous integration, testing ML pipelines, serving models through APIs, and monitoring ML systems in production. It also includes concepts around responsible AI and reproducibility. This is a true MLOps bootcamp in a box, particularly valuable to engineers working on production systems.

Best For:

  • MLOps and AI engineers deploying ML systems in the real world.
  • Teams building large-scale ML infrastructure.
  • Learners wanting a project-oriented experience of end-to-end ML.

GitHub Repository for AI Engineers: https://github.com/GokuMohandas/Made-With-ML

Hands-On Large Language Models

Hands-On LLMs provides a workflow for building and tuning large language models. The repository extends the popular O’Reilly book and features interactive notebooks that explore tokenization, attention, transformer blocks, RAG (retrieval-aided generation), embeddings, and evaluation methods. It uses Hugging Face Transformers and LangChain integrations to provide a foundation for developing real-world applications with full interpretability and modularity, such as chatbots, summarizers, and document QA systems.

Best For:

  • Engineers implementing LLMs into tangible, real-world applications.
  • Developers who will fine-tune models for specific domain tasks.
  • Researchers investigating prompt strategies and evaluation metrics.

AI-based GitHub Repository: https://github.com/pinecone-io/handbook-llms

Advanced RAG Techniques

This repository contains over 30 adaptations of the Retrieval-Augmented Generation (RAG) method, such as HyDE, GraphRAG, and complex chunking approaches. It supports experimenting with different embedding models, vector stores, document splitting, reranking, and performance benchmarking. The community can search for methods to reveal the most suitable approaches for each case, using document and query types as performance criteria, optimizing LLM-driven search and QA solutions.

Best For:

  • AI engineers designing and building RAG systems for the industry.
  • Teams trying to make the knowledge retrieval process faster while maintaining quality.
  • Scientists conducting comparative studies of vector search, hybrid, and graph approaches.

GitHub Repository: https://github.com/NirDiamant/RAG_Techniques

AI Agents for Beginners

This user-friendly repository from Microsoft introduces learners to AI agents—autonomous systems powered by LLMs that can plan, decide, and act. It features 11 experiential labs using AutoGen, LangChain, OpenAI APIs, etc., to code agents that can perform multi-step, multi-turn tasks, invoke tools, search for knowledge, and collaborate with other agents. Each lab introduces action planning, tool chaining, memory, and prompt engineering concepts clearly and reproducibly.

Best for:

  • Developers new to AI agents or agentic workflows.
  • Educators developing hands-on agent-based AI curriculums.
  • Hackers building autonomous task agents from the ground up.

GitHub LLM Repository: https://github.com/microsoft/AI-Agents

Agents Towards Production

Agents Towards Production is a comprehensive guide for transitioning AI agents from proof of concept to production. It covers implementation patterns for orchestration, tool integration, error processing, retry logic, security, memory (Redis, vector DBs), and deployment with FastAPI and Docker. Interest in scalable agentic systems is growing, and this repository serves as a template for shipping reliable and scalable agent workflows to industry.

Best For:

  • Developers deploying AI agents in production.
  • Teams building full-stack agenting infrastructure.
  • Professionals using LangGraph, OpenAgents, or AutoGen.

GitHub LLM Repository: https://github.com/NirDiamant/agents-towards-production

AI Engineering Hub

AI Engineering Hub is a vast, curated collection of 70+ real-world projects, tutorials, and templates across LLMs, RAG, and autonomous agents. It is designed for engineers wanting to enhance their skills through practical experiences. Each project includes difficulty and category tagging, with links to Colab, references, and suggested customizations. The Hub is a digital sandbox for learning every AI tool you’ve ever wanted to try, ready to fork and remix.

Best For:

  • Building a portfolio of GenAI and agent-based applications.
  • Practicing advanced LLM workflows modularly.
  • Experimenting with new tools and frameworks.

GitHub Repository: https://github.com/ashishps1/learn-ai-engineering

Conclusion

To excel in AI, you need to build and iterate with the right tools. The GitHub LLM repositories discussed provide a comprehensive package, enabling you to go from learning about machine learning to interacting with AI agents in real time. If you’re focused on deep learning, large language models (LLMs), retrieval-augmented generation (RAG), or agent orchestration, you have numerous strong real-world projects to draw on.

Explore these repositories, fork the code, modify the models, and build something unique. In the fast-paced field of AI, active learning is key, and these repositories provide an excellent way to stay active.

Frequently Asked Questions

Q1. Why should I explore GitHub repos as an AI engineer?

A. GitHub is where most cutting-edge AI work happens in public. Whether you’re learning, prototyping, or debugging, real-world code from top engineers is the best resource available.

Q2. Do I need to be an expert coder to use these repositories?

A. Not at all. Some are beginner-friendly, like ML-For-Beginners and AI-For-Beginners, which walk you through concepts with explanations and exercises.

Q3. Can I use the code from these repos in my own projects?

A. Yes, in most cases. Check the license of each repository; most are open-source under MIT or Apache licenses, which are permissive for personal and commercial use.

Q4. What’s the difference between “AI for Beginners” and “ML for Beginners”?

A. “ML for Beginners” focuses mainly on machine learning concepts like regression or classification. “AI for Beginners” is broader, including NLP, computer vision, and even ethics in AI.

Q5. Which repo is best if I want to learn how large language models work?

A. Check out nn-zero-to-hero by Andrej Karpathy. It provides one of the most hands-on and clear breakdowns of how transformers and LLMs work from scratch.

Q6. How do I keep track of updates in these repositories?

A. You can “watch” the repository on GitHub to get notifications or star it to bookmark it. You can also follow the repository maintainers if you’re particularly interested in their work.

 

Scroll to Top