Awesome Ai Learning

Awesome AI Learning Resources

[](https://awesome.re) [](https://github.com/Ming-H/awesome-ai-learning/stargazers) [](https://github.com/Ming-H/awesome-ai-learning/commits/main) [](./LICENSE)

400+ curated resources across 12 categories / 400+ 精选资源，覆盖 12 大方向

AI learning resources curated for developers and researchers, from prompt engineering fundamentals to advanced AI agent development, RAG systems, LLM inference deployment, and the path toward AGI.

为开发者和研究者精选的 AI 学习资源合集，从提示工程基础到高级 AI Agent 开发、 RAG 系统构建、大模型推理部署，直至通向 AGI 的前沿研究。

[English](#contents) | [中文](#分类目录)

---

Contents

• [Trending / What's New](#trending--whats-new)
• [Beginner / 入门](#beginner--入门)
• [01 Prompt Engineering](#01-prompt-engineering)
• [02 LLM Fundamentals](#02-llm-fundamentals)
• [03 AI Coding Tools](#03-ai-coding-tools)
• [Intermediate / 进阶](#intermediate--进阶)
• [04 RAG](#04-rag)
• [05 AI Agent](#05-ai-agent)
• [06 Skills / Function Calling](#06-skills--function-calling)
• [07 MCP](#07-mcp)
• [12 LLM Inference & Deployment](#12-llm-inference--deployment)
• [Advanced / 高级](#advanced--高级)
• [08 Claude Code](#08-claude-code)
• [09 AI Safety & Alignment](#09-ai-safety--alignment)
• [10 AGI](#10-agi)
• [11 Multimodal AI](#11-multimodal-ai)
• [Learning Path](#learning-path)
• [Contributing](#contributing)
• [License](#license)

---

Trending / What's New

• 2026-06: Added Context Engineering section (survey paper, Anthropic/LangChain guides, 6+ tools/frameworks) — "Prompt Engineering 2.0"
• 2026-06: Added Agent Guardrails section (Forge, NeMo Guardrails, Guardrails AI, Sponsio, AgentDoG) — 护栏比模型规模更重要
• 2026-06: Expanded A2A Protocol with ecosystem tools (Waggle, A2Apex, DeepLearning.AI course, ADK 2.0)
• 2026-06: Added 2026 Industry Reports (Stanford AI Index, State of AI Agents, HBR AI usage survey)
• 2026-05: New 12 LLM Inference & Deployment category (vLLM, PagedAttention, batch inference, production observability)
• 2026-05: Expanded Inference Optimization: vLLM deployment, batch inference (Ray Data), PagedAttention deep dive, 25+ new resources
• 2026-05: Added Anthropic Academy (17 free courses), Agent Memory & Evaluation sections, DeerFlow/Mastra frameworks
• 2026-05: Added MCP official courses (Anthropic Skilljar, DeepLearning.AI), DeepSeek V4, Sora, OWASP Agentic AI
• 2026-05: Added new 11 Multimodal AI category (VLM, Image/Video Gen, Multimodal RAG, Audio)
• 2026-05: Added LLM Evaluation & Benchmarking section (Chatbot Arena, HELM, HF Leaderboard, DeepEval)
• 2026-05: Added Production Deployment & Observability section (Langfuse, SGLang)
• 2026-05: Expanded Vibe Coding section (Karpathy origin, MIT Tech Review, Datawhale vibe-vibe)
• 2026-05: Added Fairness & Bias tools (IBM AIF360, Fairlearn, Google What-If Tool)
• 2026-05: Expanded 06 Skills with Structured Output, Tool Evaluation, Framework Libraries
• 2026-05: Added Microsoft GraphRAG, Agentic RAG, RAG evaluation frameworks
• 2026-05: Added Karpathy nanoChatGPT, 2025 Year in Review (RLVR), Stanford CS336, MIT 6.S191
• 2026-05: Added Anthropic-OpenAI joint alignment evaluation, AI red teaming resources
• 2026-05: Added MCP 2025-11-25 spec (Tool Annotations, OAuth), A2A Protocol
• 2026-05: Project launched with 200+ curated resources across 10 categories

---

Beginner / 入门

01 Prompt Engineering

[📁 Full Resources](./01-prompt-engineering/) · 20+ resources

提示工程教程、课程和最佳实践 / Tutorials, courses, and best practices for prompt engineering.

⭐ Editor's Picks:

• ChatGPT Prompt Engineering for Developers - Andrew Ng + OpenAI, the most recommended entry course. Free Course EN
• DAIR.AI Prompt Engineering Guide | 中文版 - The most comprehensive open-source reference. Guide EN/ZH
• dair-ai/Prompt-Engineering-Guide - 66k+ stars, the most authoritative repo. Repo EN

02 LLM Fundamentals

[📁 Full Resources](./02-llm-fundamentals/) · 50+ resources

大模型原理、微调、评估基准和分布式训练 / LLM architecture, fine-tuning, evaluation, benchmarking, and distributed training.

⭐ Editor's Picks:

• Andrej Karpathy: Neural Networks Zero to Hero - Widely considered the best starting point for deep learning and LLM fundamentals. Video Course EN
• Andrej Karpathy: nanoChatGPT (2025) - Build a complete ChatGPT clone in 8K lines of code for ~$100. Repo EN
• Sebastian Raschka: Build a LLM From Scratch - Step-by-step build, pretrain, and fine-tune a GPT-like model (50k+ stars). Book + Repo EN
• LLaMA-Factory - Zero-code fine-tuning framework for 100+ models, most popular in the CN community. Repo ZH/EN

03 AI Coding Tools

[📁 Full Resources](./03-ai-coding-tools/) · 40+ resources

Cursor、Copilot、Windsurf 等 AI 编程工具 / AI-powered coding tools: Cursor, Copilot, Windsurf, and more.

⭐ Editor's Picks:

• Cursor Official Docs - Complete official documentation for the leading AI-native IDE. Docs EN
• GitHub Copilot Official Docs - Full docs for chat, editing, agent, and CLI features. Docs EN
• 2026 AI编程工具红黑榜 - Real data and pricing comparison of 8 major tools. Article ZH

---

Intermediate / 进阶

04 RAG

[📁 Full Resources](./04-rag/) · 30+ resources

向量数据库、嵌入模型、检索策略和评估框架 / Vector databases, embeddings, retrieval strategies, and evaluation.

⭐ Editor's Picks:

• Building and Evaluating Advanced RAG - DeepLearning.AI - Sentence-window, auto-merging retrieval and more. Free Course EN
• Anthropic: Contextual Retrieval - Breakthrough method significantly improving RAG accuracy. Blog EN
• Microsoft GraphRAG - Knowledge graph-enhanced RAG for multi-hop reasoning. Repo EN
• langchain-ai/rag-from-scratch - LangChain official step-by-step RAG tutorial. Repo EN

05 AI Agent

[📁 Full Resources](./05-ai-agent/) · 45+ resources

Agent 框架、课程、架构模式、记忆和评估 / Agent frameworks, courses, architecture patterns, memory, and evaluation.

⭐ Editor's Picks:

• Agentic AI - DeepLearning.AI (Andrew Ng) - Four core agent design patterns. Free Course EN
• LangGraph Official Docs - The leading production agent orchestration framework. Docs EN
• OpenAI Agents SDK - Lightweight multi-agent framework, 19k+ stars. Docs EN
• Agentic Design Patterns (Free e-Book) - Multi-specialist agents > single large agent. PDF ZH

06 Skills / Function Calling

[📁 Full Resources](./06-skills/) · 20+ resources

工具使用、函数调用和 Agent Skills 开发 / Tool use, function calling, and agent skills development.

⭐ Editor's Picks:

• Anthropic: Writing Tools for Agents - How to design better tool descriptions. Blog EN
• Anthropic Tool Use Docs - Official Claude tool use documentation with multi-tool support. Docs EN

07 MCP

[📁 Full Resources](./07-mcp/) · 30+ resources

MCP 协议文档、官方课程、教程和服务器生态 / MCP protocol documentation, official courses, tutorials, and server ecosystem.

⭐ Editor's Picks:

• MCP Official Docs - The authoritative reference for the protocol. Docs EN
• MCP Specification 2025-11-25 - One year anniversary spec with Tool Annotations and OAuth. Spec EN
• awesome-mcp-servers - 62k+ stars, community MCP server collection. Repo EN
• Claude MCP 完全攻略 - 掘金 - Most comprehensive Chinese MCP guide (2025). Article ZH

12 LLM Inference & Deployment

[📁 Full Resources](./12-llm-inference/) · 35+ resources

大模型推理优化、vLLM 部署、批量推理和生产可观测性 / LLM inference optimization, vLLM deployment, batch inference, and production observability.

⭐ Editor's Picks:

• vLLM: Efficient Memory Management for LLM Serving (arXiv) - The original PagedAttention paper, foundation of all vLLM optimizations. Paper EN
• Introl: vLLM Production Deployment - Stripe achieved 73% inference cost reduction with vLLM. Article EN
• Anyscale: Ray Data LLM 2x Throughput - Ray Data LLM doubles throughput vs synchronous vLLM for batch inference. Article EN
• PagedAttention from First Principles - Understanding PagedAttention, the key to vLLM's performance. Article EN

---

Advanced / 高级

08 Claude Code

[📁 Full Resources](./08-claude-code/) · 20+ resources

Claude Code Skills、Hooks、Plugins 和 MCP 集成 / Claude Code skills, hooks, plugins, and MCP integration.

⭐ Editor's Picks:

• Claude Code Features Overview - Plugins bundle skills + hooks + subagents + MCP. Docs EN
• Claude Agent SDK Overview - Build autonomous agents with Claude Code's toolchain. Docs EN
• Towards AI: Complete Beginner's Guide - From installation to full feature coverage. Article EN

09 AI Safety & Alignment

[📁 Full Resources](./09-ai-safety/) · 30+ resources

AI 安全、对齐、红队测试和治理框架 / AI safety, alignment, red teaming, and governance frameworks.

⭐ Editor's Picks:

• 2026 International AI Safety Report - Second international report from the Bletchley Park AI Safety Summit. Report EN/ZH
• Anthropic-OpenAI Joint Alignment Evaluation - First-ever cross-lab safety evaluation. Research EN
• Promptfoo - Open-source CLI for LLM evaluation and red teaming. Repo EN
• NeurIPS 2025 Tutorial: Human-AI Alignment - Research-grade tutorial with videos and slides. Tutorial EN

10 AGI

[📁 Full Resources](./10-agi/) · 15+ resources

AGI 研究、时间线和政策报告 / AGI research, timelines, and policy reports.

⭐ Editor's Picks:

• Sequoia: 2026 This is AGI - Long-horizon coding agents as functional AGI. Article EN
• Andrej Karpathy: 2025 LLM Year in Review - RLVR replacing RLHF, the paradigm shift on the path to AGI. Article EN
• WaytoAGI 飞书知识库 - China's largest AI learning community (~9M learners). Community ZH
• WEF: Preparing for AGI (2025) - Comprehensive global policy document. PDF EN

11 Multimodal AI

[📁 Full Resources](./11-multimodal-ai/) · 20+ resources

视觉语言模型、图像/视频生成、多模态 RAG 和语音 AI / Vision-language models, image/video generation, multimodal RAG, and audio AI.

⭐ Editor's Picks:

• Hugging Face: Vision Language Models Explained - The canonical primer on VLMs from Hugging Face. Article EN
• ComfyUI - 70k+ stars, the most powerful modular diffusion model GUI/backend. Repo EN
• Building Multimodal Search and RAG - DeepLearning.AI - Free course on building multimodal RAG systems. Free Course EN
• OpenAI Whisper - 80k+ stars, the most powerful open-source speech recognition model. Repo EN

---

Learning Path

入门 / Beginner
  ├── 01 Prompt Engineering (Andrew Ng course)
  ├── 02 LLM Fundamentals (Karpathy Zero to Hero → nanoChatGPT → Evaluation)
  └── 03 AI Coding Tools (Cursor / Copilot / Vibe Coding)

进阶 / Intermediate
  ├── 04 RAG Systems (Vector DB → GraphRAG → Agentic RAG → Multimodal RAG)
  ├── 05 Agent Frameworks (LangGraph / OpenAI SDK / Google ADK / Memory / Evaluation)
  ├── 06 Skills / Function Calling (Structured Output, Tool Evaluation)
  ├── 07 MCP Protocol & Development (Tool Annotations, OAuth)
  └── 12 LLM Inference & Deployment (vLLM / Batch Inference / Production)

高级 / Advanced
  ├── 08 Claude Code (Skills, Hooks, Agent SDK, CLAUDE.md)
  ├── 09 AI Safety & Alignment (Cross-lab eval, Red Teaming, Fairness)
  ├── 10 AGI Research (RLVR paradigm, AAAI survey, Timelines)
  └── 11 Multimodal AI (VLM, Image/Video Gen, Audio)

---

Contributing

PRs are welcome! Please read the [Contributing Guide](./CONTRIBUTING.md) and [Code of Conduct](./CODEOFCONDUCT.md) before submitting.

Format for new entries:

- [Resource Name](URL) - Description ending with a period. `Type` `Language`

License

[MIT](./LICENSE) © Ming-H

---

Star History

[](https://star-history.com/#Ming-H/awesome-ai-learning&Date)

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Ming-H
• Source: Ming-H/awesome-ai-learning
• License: MIT

Install and usage instructions live in the source repository linked above.