Photo by author
# Introduction
Learning AI today is not just about understanding machine learning models. It’s about knowing how things fit together in practice, from math and fundamentals to building real applications, agents, and production systems. With so much content online, it’s easy to feel lost or jump between random lessons without a clear path.
In this article, we will learn about 10 popular and really useful GitHub repositories for learning AI. These repos cover the full spectrum, including generative AI, large language models, agentic systems, mathematics for ML, computer vision, real-world projects, and production-grade AI engineering.
# GitHub repositories for learning AI
// 1. Official for Microsoft/Generative-AI
Generative AI for beginners A 21-lesson course created by Microsoft Cloud Advocates that teaches how to build real generative AI applications from scratch. It combines clear conceptual lessons with hands-on builds in Python and TypeScript, covering hints, chat, rags, rigs, agents, fine-tuning, security, and deployment. The course is beginner-friendly, multilingual, and designed to move learners from fundamentals to production-ready AI apps with practical examples and community support.
// 2. rasbt/llms-from-scratch
Build a large language model (from scratch) Manning’s book is a hands-on, educational resource and companion that teaches how LLMS works by implementing a step-by-step GPT-style model in pure pytorch. It runs through tokenization, focus, GPT architecture, pretraining, and fine-tuning (including instruction tuning and Laura), all designed to run on a regular laptop. The focus is on deep understanding through code, diagrams and exercises rather than using high-level LLM libraries, making it ideal for learning LLM internals from the ground up.
// 3. Data Talks Club/LLM Zumcamp
LL.M. Zumkamp is a free, 10-week course focused on building real-world LLM applications, specifically RAG-based systems on your own data. It covers vector detection, assessment, surveillance, agents and best practices through practical workshops and capstone projects. Designed for self-paced or synchronous learning, it emphasizes production-ready skills, community feedback, and building end-to-end systems rather than just theory.
// 4. Shubhamsabu/Horrible LLM apps
Terrible LLM apps RAG is a curated showcase of real, runnable LLM applications built with AI agents, multi-agent teams, MCP, voice interfaces, and memory. It highlights practical projects using OpenAI, Entropic, Gemini, Zee, and open source models such as Llama and Quen, many of which can be run natively. The focus is on learning by example, exploring modern agent patterns, and accelerating the development of production-style LLM apps.
// 5. Panaversity/Learning Ai
Learn agentic AI using DAPR Agentic Cloud Ascent (DACA). A cloud-native, systems-first learning program focused on designing and scaling agentic AI systems on a planetary scale. It teaches how to build reliable, interoperable multi-agent architectures using Kubernetes, DAPR, the OpenAI Agent SDK, MCP, and the A2A protocol, with a strong emphasis on workflow, flexibility, cost control, and real-world execution. Its purpose is not just to build agents, but to train developers to design production-ready agent swarms that can scale to millions of concurrent agents under realistic constraints.
// 6. ML for Der-A/Maths
Mathematics for Machine Learning A curated collection of high-quality books, papers, and video lectures covering the mathematical foundations behind modern ML and deep learning. It focuses on core areas such as linear algebra, calculus, probability, statistics, optimization and information theory, with resources ranging from beginner-friendly to research-level in-depth. It aims to help learners develop a strong mathematical intuition and confidently understand the theory behind machine learning models and algorithms.
// 7. Asheshiptel 26/500-AI-Machine-learning-deep-learning-learning-vision-nlp-projects-ath-code
List of 500+ artificial intelligence projects with code AI/ML/DL is a large, continuously updated directory of project ideas and learning resources, grouped into areas such as computer vision, NLP, time series, recommender systems, healthcare, and production ML. It links to hundreds of tutorials, datasets, GitHub repos, and “projects with source code” and encourages community contributions through pull requests to enhance links and collections.
// 8. Ermankhandkar/horrible-i-ml-resources
Machine Learning and AI Roadmap (2025) A structured, beginner-to-advanced guide that maps out learning AI and machine learning step-by-step. It covers basic concepts, mathematical foundations, tools, roles, projects, MLOPS, interviews and research, while linking to trusted courses, books, papers and communities. It aims to give learners a clear path into a fast-paced field, helping them develop practical skills and career preparation without overwhelming them.
// 9. spmallick/learnopencv
Learn Open There is a comprehensive, hands-on repository that accompanies the LearnOpenCo.com blog, featuring hundreds of tutorials with runnable code in computer vision, deep learning, and advanced AI. It spans topics from classic OpenCV fundamentals to cutting-edge models such as ULO, SAM, diffusion models, VLMS, robotics, and AAI, with a strong focus on practical implementation. The collection is ideal for learners and practitioners who want to understand AI concepts by building real systems rather than just reading theory.
// 10. x1xhlol/system-prompts and model-of-AI tools
System notation and models of AI tools is an open-source AI engineering repository that documents how real-world AI tools and agents are structured, exposing more than 30,000 lines of system notation, model behavior, and design patterns. It is particularly useful for developers building trusted agents and indicators, offering practical insight into how production AI systems are designed, while also highlighting the importance of immediate security and leak prevention.
# Final thoughts
From my experience, the fastest way to learn AI is to stop treating it as a theory and start building as you learn. These repositories work because they are built by real engineers who are practical, opinionated, and solving real problems.
My advice is to pick a few that match your current level and goals, get to the end of them, and build steadily. Depth, repetition, and more depth, fulfillment than chasing every new trend.
Abid Ali Owan For centuries.@1abidaliawan) is a certified data scientist professional who loves building machine learning models. Currently, he is focusing on content creation and writing technical blogs on machine learning and data science technologies. Abid holds a Master’s degree in Technology Management and a Bachelor’s degree in Telecommunication Engineering. His vision is to create an AI product using graph neural networks for students with mental illness.