A structured, tutorial-style knowledge base covering Large Language Models, Retrieval-Augmented Generation, and AI Agents — from foundations to advanced architectures.
graph LR
A[LLM Foundations] --> B[Hardware & Infrastructure]
A --> C[Retrieval-Augmented Generation]
C --> D[AI Agents]
A --> A1[How LLMs Work]
A --> A2[Training & Fine-Tuning]
B --> B1[GPU Architecture]
B --> B2[Infrastructure & Scaling]
B --> B3[Energy Efficiency]
C --> C1[Chunking Strategies]
C --> C2[Context Engineering]
C --> C3[Vector Databases]
D --> D1[Agent Fundamentals]
D --> D2[Multi-Agent Architectures]
style A fill:#4a90d9,stroke:#2c5f8a,color:#fff
style B fill:#f39c12,stroke:#c87f0a,color:#fff
style C fill:#7b68ee,stroke:#5a4cb5,color:#fff
style D fill:#e06c75,stroke:#a84e55,color:#fff
style A1 fill:#6aacf0,stroke:#4a8ad0,color:#fff
style A2 fill:#6aacf0,stroke:#4a8ad0,color:#fff
style B1 fill:#f5a962,stroke:#d68e4a,color:#fff
style B2 fill:#f5a962,stroke:#d68e4a,color:#fff
style B3 fill:#f5a962,stroke:#d68e4a,color:#fff
style C1 fill:#9a8aee,stroke:#7a6ace,color:#fff
style C2 fill:#9a8aee,stroke:#7a6ace,color:#fff
style C3 fill:#9a8aee,stroke:#7a6ace,color:#fff
style D1 fill:#e89a9f,stroke:#c87a7f,color:#fff
style D2 fill:#e89a9f,stroke:#c87a7f,color:#fff
- Section Overview
- How LLMs Work — Evolution from statistical models to modern LLMs, Transformer architecture, emergent abilities
- Training & Fine-Tuning — Pre-training, SFT, RLHF, DPO, and practical considerations
- Section Overview
- GPU Architecture for LLMs — Why GPUs excel at transformers, tensor cores, memory hierarchy, and GPU evolution (Volta → Ampere → Hopper)
- Infrastructure & Scaling — Scaling laws, data center requirements, distributed training challenges, cost analysis, and the infrastructure arms race
- Energy Efficiency — Power consumption, environmental impact, hardware and algorithmic optimizations, sustainability approaches
- Section Overview
- Chunking Strategies — How to split documents for optimal retrieval, based on Chroma research
- Context Engineering — Why context degrades and how to fight it, based on Chroma research
- Vector Databases — Elasticsearch vs. Qdrant vs. Milvus comparison
- Section Overview
- Agent Fundamentals — ReAct, tool use, memory, planning
- Memory and State Management — Conversation memory, long-term recall, episodic memory, state machines, checkpointing
- Multi-Agent Architectures — Patterns, LangChain, LangGraph
- Key Papers — Curated reading list with annotations
- Tools & Frameworks — Reference guide to the ecosystem
- Hardware for LLMs — GPUs, TPUs, memory, cost analysis, optimization
- ML/AI Engineers who want a structured reference on LLM internals, RAG pipelines, and agent architectures
- Software Engineers transitioning into AI/ML who need practical, concept-first explanations
- Technical Leaders evaluating where LLMs and agents fit into their stack
- Students & Researchers looking for a curated, opinionated guide through the literature
| Your Goal | Start Here |
|---|---|
| Understand how LLMs work from scratch | How LLMs Work |
| Understand GPU architecture and scaling | GPU Architecture for LLMs |
| Evaluate infrastructure requirements | Infrastructure & Scaling |
| Optimize energy efficiency | Energy Efficiency |
| Build or improve a RAG pipeline | Chunking Strategies |
| Choose a vector database | Vector Databases |
| Understand long context and when to use it vs. RAG | Long Context and Context Windows |
| Understand AI agents | Agent Fundamentals |
| Design agent memory systems | Memory and State Management |
| Design multi-agent systems | Multi-Agent Architectures |
| Find key papers to read | Papers |
| Choose hardware for training/inference | Hardware for LLMs |
Want to add a topic or improve existing content? See the Contributing Guide for templates, style conventions, and how to submit changes.