title

AI Guru Knowledge Base (English)

AI Guru Knowledge Base

Probably the Most Comprehensive AI Learning Resource on GitHub

Complete AI Knowledge System from Theory to Production | 2,061 Core Docs + 700+ Concept Cards | 18.9M Characters | LLMOps End-to-End | 2026 Latest

Quick Start • Why AI Guru • Content Navigation • Who Is It For • Download for Local Tools

中文版

Why AI Guru?

The Problem: AI evolves faster than ever. Learners struggle with fragmented information, outdated content, and the widening gap between theory and production deployment.

The Solution: AI Guru is a production-grade knowledge system that bridges theory and practice, taking you from zero to AI mastery with a complete LLMOps pipeline.

Key Highlights

** Comprehensive Content**

2,061 core docs + 700+ atomic concept cards
~18.9M characters (~3,000 A4 pages)
24 knowledge chapters + 12 subdomain concept network
From math foundations to AGI frontiers

** Dual Learning Tracks**

Fast Path: 70+ in-nutshell guides (hands-on)
Systematic: 24 in-depth chapter deep-dives
Beginner-Friendly: 48 for_dummy guides
Career Prep: 140 interview guides

** 2026 Latest**

GPT-5.2 / Claude 4.5 / DeepSeek-R1 architectures
China's top 6 LLM vendors (DeepSeek/Qwen/GLM/Kimi/MiniMax/MiMo)
LLMOps end-to-end (Prompt/Eval/RAG/Cost/Observability)
Agent production deployment + MCP/A2A protocol

** Production-Ready**

K8s + GPU cluster deployment blueprints
MLOps + LLMOps full pipeline
Inference cost optimization (quant/routing/caching)
Enterprise security & compliance (EU AI Act)

By The Numbers

 2,061 Core Docs         18.9M chars (~3,000 A4 pages)
 24 Knowledge Chapters   70+ Quick Guides (in-nutshell)
 48 Beginner Guides (for_dummy)   700+ Concept Cards (12 subdomains)
 240 Agent Articles      157 Industry Leader Perspectives
 140 Interview Guides    87 Industry Case Studies

Core docs (24 chapter directories): 2,061 files / 18.9M chars Concept cards (12 subdomains): 700+ files

Per-Directory Stats

Directory	Files	Characters	Share
概念 (Concepts)	700+	—	—
15_智能体 (Agents)	240	2.84M	15.0%
19_业界观点 (Insights)	157	960K	5.1%
05_大模型 (LLMs)	154	2.30M	12.2%
21_面试岗位 (Interviews)	140	875K	4.6%
12_架构基建 (Infrastructure)	140	1.51M	8.0%
90_学习 (Learning)	123	862K	4.6%
11_模型运维 (MLOps)	115	1.13M	6.0%
10_部署推理 (Deployment)	105	940K	5.0%
16_编程 (AI Coding)	95	655K	3.5%
18_行业应用 (Industry)	87	420K	2.2%
22-FDE (Full-Stack Eng)	71	254K	1.3%
20_论文精读 (Papers)	69	585K	3.1%
02_机器学习 (ML)	64	455K	2.4%
治理 (Governance)	64	—	—
17_伦理安全 (Safety)	63	438K	2.3%
07_模型训练 (Training)	62	763K	4.0%
14_RAG系统 (RAG)	62	567K	3.0%
01_数学基础 (Math)	60	492K	2.6%
03_深度学习 (DL)	55	499K	2.6%
13_运维 (SRE/Ops)	52	375K	2.0%
08_模型评估 (Evaluation)	48	585K	3.1%
06_强化学习 (RL)	44	535K	2.8%
04_计算机视觉 (CV)	41	313K	1.7%
09_测试 (Testing)	30	318K	1.7%
00_入门 (Intro)	27	242K	1.3%
94_可视化 (Visualization)	28	271K	1.4%
Total	2,061	18.9M	100%

Tip: Run python3 工具/count_words.py for real-time stats.

Quick Start

Option 1: Read Directly (Recommended)

# Clone the repository
git clone https://github.com/your-org/ai-guru-knowledge-base.git

# Navigate to root
cd ai-guru-knowledge-base

# Start from README
less README.md

Option 2: Web Interface

cd 前端应用/
npm install
npm run dev
# Visit http://localhost:5173

Option 3: Import to AI Tools

Supports NotebookLM, ima, Claude Projects, etc.:

# Download complete knowledge base
git clone --depth 1 https://github.com/your-org/ai-guru-knowledge-base.git

Core chapter structure: Each chapter has a README.md entry point + for_dummy beginner guide + in-nutshell quick reference — a three-tier reading path.

Who Is It For?

** Ops/Dev Engineers**

Transition to AI Agent Engineer

15-20 hours

Fast Path →

** Students/Self-Learners**

Master AI full-stack systematically

16-20 weeks

University Course →

** Product Managers**

Understand AI capabilities & boundaries

8-10 hours

Industry Apps →

** Researchers**

Track cutting-edge technology

Self-paced

Essential Papers →

Content Navigation

Knowledge Graph & Learning Path

graph TD
    %% ===== Layer 0: Entry =====
    ENTRY[" Entry<br/>AI Fundamentals"]
    MATH[" Math<br/>Linear Algebra · Prob · Calculus · Info Theory"]
    CONCEPT[" Concepts<br/>700+ Term Cards · 12 Subdomains"]
    PROGRAMMING[" Programming<br/>AI Coding · OpenCode · Vibe Coding"]

    %% ===== Layer 1: Theory =====
    ML[" Machine Learning<br/>Supervised · Unsupervised · Ensemble"]
    DL[" Deep Learning<br/>Neural Nets · Optimization · GAN/VAE/Diffusion"]
    RL[" Reinforcement Learning<br/>MDP · DQN/PPO · SAC · RLHF"]
    CV[" Computer Vision<br/>Detection · Segmentation · 3D · Video Gen"]
    NLP[" NLP/Sequence Models<br/>RNN · LSTM · Tokenizer"]

    %% ===== Layer 2: LLM Core =====
    LLM[" LLMs<br/>Transformer · MoE · Fine-tuning · Reasoning"]
    MULTIMODAL[" Multimodal<br/>Vision-Language · Speech · Video"]
    PAPERS[" Papers<br/>Transformer · BERT · GPT · RLHF · DPO"]

    %% ===== Layer 3: Training & Eval =====
    TRAINING[" Training<br/>Distributed · Data Eng · Alignment · Compression"]
    EVAL[" Evaluation<br/>Benchmark · LLM-as-Judge · Red Team"]
    SAFETY[" Safety<br/>Alignment · Guardrails · RSI · Supply Chain"]

    %% ===== Layer 4: Deployment & Ops =====
    DEPLOY[" Deployment<br/>vLLM/SGLang · Quantization · KV Cache"]
    ARCH[" Architecture<br/>K8s · GPU Virt · AI Gateway · DR"]
    OPS[" MLOps/LLMOps<br/>Observability · CI/CD"]
    SRE[" SRE<br/>Troubleshooting · Capacity · Chaos · FinOps"]

    %% ===== Layer 5: Application =====
    RAG[" RAG Systems<br/>Vector DBs · Hybrid · Agentic RAG"]
    AGENT[" Agents<br/>Frameworks · Eval · Harness · Protocols"]
    INDUSTRY[" Industry<br/>Finance · Healthcare · Manufacturing"]
    INTERVIEW[" Interviews<br/>25+ Roles · System Design"]
    FDE[" Full-Stack Eng<br/>End-to-End Projects"]
    VIZ[" Visualization<br/>Dashboards · Explainability"]
    VIEWS[" Perspectives<br/>28 Leaders · Keynotes"]
    TEST[" Testing<br/>RAGAS · DeepEval · Contract Testing"]
    LEARN[" Learning<br/>Course Maps · Career Paths"]

    %% ===== Edges =====
    ENTRY --> MATH & PROGRAMMING & CONCEPT
    MATH --> ML & DL
    PROGRAMMING --> ML
    ML --> DL & RL & CV
    DL --> NLP & MULTIMODAL
    NLP --> LLM
    DL --> LLM
    CONCEPT -.->|Reference| ML & DL & LLM & RL
    LLM --> MULTIMODAL & PAPERS
    RL -->|RLHF/DPO/GRPO| LLM
    LLM --> TRAINING & DEPLOY
    TRAINING --> EVAL
    EVAL --> SAFETY
    PAPERS -.->|Theory| TRAINING & EVAL
    DEPLOY --> ARCH & OPS
    ARCH --> SRE
    OPS --> SRE
    LLM --> RAG
    RAG --> AGENT
    LLM --> AGENT
    AGENT --> INDUSTRY & FDE
    DEPLOY --> INDUSTRY
    EVAL --> TEST
    EVAL --> VIZ
    ENTRY --> LEARN
    LEARN --> INTERVIEW
    LLM --> VIEWS

    %% ===== Styling =====
    classDef entry fill:#4CAF50,stroke:#2E7D32,color:#fff
    classDef math fill:#2196F3,stroke:#1565C0,color:#fff
    classDef core fill:#9C27B0,stroke:#6A1B9A,color:#fff
    classDef llm fill:#FF9800,stroke:#E65100,color:#fff
    classDef deploy fill:#F44336,stroke:#C62828,color:#fff
    classDef app fill:#00BCD4,stroke:#00838F,color:#fff
    classDef meta fill:#607D8B,stroke:#37474F,color:#fff

    class ENTRY entry
    class MATH,CONCEPT,PROGRAMMING math
    class ML,DL,RL,CV,NLP,MULTIMODAL,PAPERS core
    class LLM,TRAINING,EVAL,SAFETY llm
    class DEPLOY,ARCH,OPS,SRE deploy
    class RAG,AGENT,INDUSTRY,INTERVIEW,FDE app
    class VIZ,VIEWS,TEST,LEARN meta

Learning Path Quick Reference:

Role	Recommended Path	Est. Time
Beginner	Intro → Math → ML → Deep Learning	16-20 weeks
LLM Engineer	LLMs → Training → Deployment → RAG → Agents	8-12 weeks
Ops/SRE	Deployment → Architecture → MLOps → SRE	4-8 weeks
AI Safety Eng	Ethics/Safety → Eval/Red Team → Concepts/Safety	4-6 weeks
Interview Prep	Interviews → Papers → Concepts → Learning	2-4 weeks
Full-Stack Eng	Intro → Coding → LLMs → FDE	4-6 weeks

University Curriculum (Entry)

Complete AI general education supporting 16-week semesters:

Module	Content	Duration
AI Fundamentals	Definitions, types, core technologies	2-3h
Technology Landscape	Tech stack, algorithms, tools	3-4h
History Timeline	1950-2026, 4 AI waves	2-3h
Tools Practice	ChatGPT, Claude, Cursor	3-4h
Ethics & Society	Bias, privacy, governance	3-4h
Future Trends	AGI path, RSI, 2026-2040	2-3h
+ Glossary + Cases + Labs	700+ terms, 6 cases, 8 labs	—

Fast Path (70+ In-Nutshell Guides)

Hands-on guides for engineers getting up to speed quickly:

Phase 1: Foundation
├── ① LLM Basics (Token, context, Temperature)
└── ② Prompt Engineering (CoT, Few-shot)

Phase 2: Core Skills
├── ③ Model Training (Loss functions, optimizers)
├── ④ Model Inference (Deployment, INT8/FP8)
└── ⑤ RAG Systems (Vector DBs, hybrid search)

Phase 3: Agent Engineering
├── ⑥ AI Agents (ReAct, Function Calling)
├── ⑦ AI Skills (Skill registration, composition)
└── ⑧ Agent Workflows (LangGraph, error handling)

Phase 4: Production
├── ⑨ AI Testing & Evaluation (Metrics, LLM-as-Judge)
├── ⑩ MLOps Pipeline (CI/CD, model registry)
└── ⑪ AI Operations (Observability, incident response)

LLMOps End-to-End Pipeline

This knowledge base's signature LLMOps end-to-end pipeline covers every critical aspect of MLOps in the LLM era (11_模型运维/ — 115 docs):

graph LR
    L[LLMOps 2026 Pipeline] --> P[Prompt Engineering Ops<br/>Versioning · A/B Testing · CI Gates]
    L --> E[LLM Eval Pipeline<br/>LLM-as-Judge · Human Review · Eval-Driven Dev]
    L --> R[RAG Pipeline Ops<br/>4D Variability · Embedding Upgrade · Index Rebuild]
    L --> C[Cost & Latency SLO<br/>3-Tier Cache · Smart Routing · Token Circuit Breaker]
    L --> O[LLM Observability<br/>5-Layer Monitoring · Trace · Hallucination/PII Detection]
    L --> G[AI Governance & Compliance<br/>EU AI Act · Audit · Model Cards]

Doc	Content
LLMOps 2026	Why traditional MLOps fails (7 reasons), 3-tier architecture, maturity model
Prompt Engineering Ops	Prompt-as-code, versioning, A/B testing, DSPy auto-optimization
LLM Eval Pipeline	LLM-as-Judge, human review loop, eval pitfalls
RAG Pipeline Ops	4D variability, embedding upgrade migration, index rebuild canary
LLM Cost & Latency SLO	3-tier cache, smart routing, token budget circuit breaker, FinOps
LLM Observability	5-layer monitoring, distributed tracing, hallucination/toxicity/PII detection

Plus cross-cutting concerns (data versioning/retraining/system SLO/cost/compliance) and 16 deep-dive tool analyses — 115 docs total for LLMOps.

24 Systematic Chapters

Complete path from math to production:

Click to View Full Directory

Chapter	Core Content	Difficulty
L0 Foundation
00 AI Intro & History	General education: concepts, landscape, history, tools, ethics
01 Fundamentals	Math & CS: Linear algebra, probability, data structures, distributed systems, AI hardware 2026
L1 Models
02 Classical ML	ML basics: Supervised/unsupervised learning, feature engineering, XGBoost
03 Deep Learning	Neural networks: MLP, backprop, optimization, JEPA world models
04 Computer Vision	Vision AI: CNN, YOLO, diffusion, video generation 2026
05 NLP & LLMs	LLM tech: Transformer, GPT-5.2/Claude 4.5, DeepSeek-R1, LoRA/RLHF/DPO
06 RL & Agents	RL & Agents: DQN/PPO, GRPO, VLA embodied AI
L2 Engineering
07 Model Training	Training: Loss functions, optimizers, distributed training, GRPO alignment
08 Model Evaluation	Evaluation: Metrics, benchmarks, LLM-as-Judge, A/B testing
09 AI Testing	Testing: Ragas/DeepEval/Promptfoo, contract testing, Eval-Driven Dev
10 Deployment & Inference	Inference: vLLM/SGLang, quantization, KV Cache, reasoning models
L3 Platform
11 MLOps Pipeline	LLMOps end-to-end + traditional MLOps + deep-dive tool analysis (115 docs)
12 Architecture & Infra	System design: K8s, multi-tenant, HA, GPU cluster, capacity planning
13 AI Ops	AIOps: Incident response, SRE practices, chaos engineering, K8s troubleshooting
L4 Application
14 RAG Systems	Retrieval augmented: Vector DBs, hybrid search, Agentic RAG, multimodal retrieval
15 Agent Production	Agent engineering: Frameworks, MCP/A2A protocol, Harness, skills, workflows, eval (240 docs)
Agent Skills	Skill system: Registration, composition, ecosystem
Agent Workflow	Workflows: LangGraph, error handling, orchestration
Agent Eval	Evaluation: Benchmark, red team, leaderboard
16 AI Coding	Coding tools & methodology: Cursor, Claude Code, Vibe Coding
22 FDE Full-Stack Eng	End-to-end full-stack projects: System design, architecture, production deployment
L5 Governance
17 Ethics & Safety	AI safety: Alignment, red team, RSI, privacy protection, OWASP LLM
L6 Resources
18 Industry Applications	Verticals: Healthcare/Finance/Manufacturing/Retail/Education/Autonomous Driving
19 Industry Perspectives	Leader insights: 28 AI pioneers, including 2026 trend collections
20 Essential Papers	Classics: Transformer, BERT, GPT, RLHF, DPO and other milestones
21 Interviews & Careers	Career: 25+ AI role interview guides, question banks, system design
Extras
90 Learning Resources	Course maps: Microsoft/Datawhale/HuggingFace courses, career paths
94 Visualization	Knowledge graph viz: Dashboards, training monitoring, eval visualization

Concept Card System (700+ Cards, 12 Subdomains)

Atomic concept cards with a "one-sentence understanding + 2026 ecosystem + production best practices" standard:

Subdomain	Cards	Description
General	151	Cross-domain concepts (basics/ML/DL/tools/cloud-native)
LLM	118	LLM architecture, training, alignment, inference
K8s	106	Kubernetes & cloud-native AI infrastructure
Training	61	Model training, distributed training, optimization
Inference	43	Inference engines, serving, performance optimization
RAG	41	Retrieval-augmented generation, vector databases
Agent	39	AI agents, tool calling, multi-agent
GPU	34	GPU hardware, CUDA, cluster management
Vision	28	Computer vision, multimodal
MLOps	28	ML ops, CI/CD, monitoring
Safety	26	AI safety, alignment, ethics, governance
Math	22	Math foundations, optimization theory
Total	700+	12 subdomains, atomic concept network

Each concept card includes: YAML frontmatter (title/aliases/tags/summary/relationships/sources), one-sentence understanding, core concept table, 2026 ecosystem table, production best practices, and further reading.

2025-2026 AI Trends Overview

These trends help practitioners quickly grasp the direction of the field.

Trend	Key Insight	Corresponding Chapter
Agentic AI at Scale	Gartner predicts ~50% of enterprise software will embed autonomous Agent tools by 2026; McKinsey estimates $2.6-4.4T annual value; but >40% of Agentic AI projects may be cut for governance gaps	15_智能体 — 240 Agent full-stack docs
Reasoning Models & Test-Time Compute	DeepSeek-R1 pioneered "test-time compute scaling"; RL training compute grows 10x every few months; o3/R1/Llama-4 reasoning model roadmap	05_大模型/Reasoning_Models + GRPO
Small Models & Efficient Inference	Compact LMs projected to handle 60% of business AI tasks, cutting infra costs 50%; "efficiency is the new scaling law"	Quantization + Edge LLM
Multimodal + Embodied AI	Multimodal models fuse vision, language and action; 58% of enterprises already use physical robots; VLA is the new frontier	Computer Vision + VLA
AI Safety & Compliance	EU AI Act enters enforcement; 80% of regulated industries must implement mandatory compliance frameworks; RSI becomes new alignment focus	Ethics & Safety + RSI Concept
MCP/A2A Protocol Standardization	MCP (Model Context Protocol) becomes the de facto Agent tool-calling standard; A2A (Agent-to-Agent) drives multi-agent collaboration	Agent Protocols

Recommendation: 2026 AI practitioners should focus on Agent engineering + Inference efficiency + Safety & compliance — this knowledge base covers all three through LLMOps + Agent + Safety.

Download for Local Tools

AI Guru can be used as high-quality corpus for various AI tools:

NotebookLM (Google)

Visit notebooklm.google.com
Create new project, select "GitHub" or upload ZIP
Paste repo URL: https://github.com/your-org/ai-guru-knowledge-base
NotebookLM automatically analyzes all documents

ima (Tencent)

Open ima app, create a knowledge base
Import local folder, select the downloaded repo root
Query via conversational interface

Claude Projects / ChatGPT GPTs

# Download lite version (core content only)
git clone --depth 1 --filter=blob:none --sparse https://github.com/your-org/ai-guru-knowledge-base.git
cd ai-guru-knowledge-base
git sparse-checkout set 00_入门 05_大模型 11_模型运维 15_智能体 概念

# Package for upload
zip -r ai-guru-core.zip 00_入门 05_大模型 11_模型运维 15_智能体 概念

Obsidian

All documents are Markdown, directly usable as an Obsidian Vault (the repo includes .obsidian/ config with wiki-link support and graph view):

# In Obsidian: "Open local folder" → select the repo root

Agent-Friendly Design

This knowledge base is optimized for AI Agent consumption:

Structured Metadata: Every doc has frontmatter (title/category/tags/summary/relationships)
Clear Hierarchy: Chapter-doc-paragraph structure with numbered directories for easy retrieval
Three-Tier Entry: Each chapter has README + for_dummy + in-nutshell
Executable Code: Runnable code examples throughout
Bilingual: Technical terms preserved in English for concept clarity
Wiki Links: 6,900+ internal wikilinks forming a knowledge graph
Concept Dictionary: 700+ atomic concept cards across 12 subdomains
Version Control: Git history for traceable updates

Recommended Agent Usage:

Import the entire repo root as a knowledge base
Use file paths as context references (e.g., 05_大模型/04_LLM架构/05_LLM架构.md)
Combine with chapter READMEs for quick content location
Use 概念/ for concept lookup and 治理/ for cross-domain linking
Use 概念/index.md as the entry index

Frontend Project

A modern knowledge base frontend (React + Vite + TypeScript + Tailwind CSS + shadcn/ui):

cd 前端应用/
npm install
npm run dev

Features:

Vite fast builds
Tailwind CSS + shadcn/ui
Full-text search (Fuse.js)
Dark/light mode
Responsive design
Agent evaluation dashboard integration

View Frontend Details →

Contributing

We welcome all forms of contributions!

How to Contribute

Content: Add new guides, update outdated information
Translation: Translate to English/other languages
Frontend: Improve the Web experience (React/Vite)
Issue: Report problems or suggestions

Quick Start

# Fork and clone
git clone https://github.com/your-username/ai-guru-knowledge-base.git

# Create branch
git checkout -b feature/your-feature

# Submit changes
git commit -m "feat: add XX content"
git push origin feature/your-feature

View Full Contributing Guide →

License

MIT License — See LICENSE for details.

Referenced papers, books, and third-party projects follow their original licenses.

Acknowledgments

Thanks to all contributors to this knowledge base.

Special thanks to:

DeepLearning.AI — Course structure inspiration
Hugging Face — Open-source ML ecosystem
All contributors and readers

AI Guru — Making AI Learning Systematic, Efficient, and Accessible

⭐ Star this repo for updates • Watch for new content • Fork to create your version

[[治理/ROADMAP.md]] — AI Guru Roadmap
[[治理/KNOWN_ISSUES.md]] — Known Issues
[[README]] — Chinese Version
[[概念/index]] — Concept Graph Home
[[概念/README]] — Concept Cards Index
[[治理/README]] — Cross-domain Index
[[治理/hot.md]] — Hot Pages
[[治理/plan]] — Planning & Evaluation
[[00_入门/01_基础入门/02_AI基础.md]] — Knowledge Base Entry

Navigation Menu

Uh oh!

AI Guru Knowledge Base

Why AI Guru?

Key Highlights

By The Numbers

Per-Directory Stats

Quick Start

Option 1: Read Directly (Recommended)

Option 2: Web Interface

Option 3: Import to AI Tools

Who Is It For?

Content Navigation

Knowledge Graph & Learning Path

University Curriculum (Entry)

Fast Path (70+ In-Nutshell Guides)

LLMOps End-to-End Pipeline

24 Systematic Chapters

Concept Card System (700+ Cards, 12 Subdomains)

2025-2026 AI Trends Overview

Download for Local Tools

NotebookLM (Google)

ima (Tencent)

Claude Projects / ChatGPT GPTs

Obsidian

Agent-Friendly Design

Frontend Project

Contributing

How to Contribute

Quick Start

License

Acknowledgments

Related

Uh oh!

FilesExpand file tree

README_EN.md

Latest commit

History

README_EN.md

File metadata and controls

AI Guru Knowledge Base

Why AI Guru?

Key Highlights

By The Numbers

Per-Directory Stats

Quick Start

Option 1: Read Directly (Recommended)

Option 2: Web Interface

Option 3: Import to AI Tools

Who Is It For?

Content Navigation

Knowledge Graph & Learning Path

University Curriculum (Entry)

Fast Path (70+ In-Nutshell Guides)

LLMOps End-to-End Pipeline

24 Systematic Chapters

Concept Card System (700+ Cards, 12 Subdomains)

2025-2026 AI Trends Overview

Download for Local Tools

NotebookLM (Google)

ima (Tencent)

Claude Projects / ChatGPT GPTs

Obsidian

Agent-Friendly Design

Frontend Project

Contributing

How to Contribute

Quick Start

License

Acknowledgments

Related