Gen-AI-Learning: Notes on Generative AI Learning and Practical Exploration

This repository records the author's learning journey and practical exploration in the field of generative AI, focusing on core technologies such as natural language processing (NLP), large language models (LLM), LangChain, and LangGraph, providing a reference path for generative AI learners.

Generative Artificial Intelligence (Generative AI) is reshaping the landscape of the technology industry. From ChatGPT attracting global attention to the commercialization of various AI applications, it has become the most transformative technological wave at present. For technical practitioners, mastering generative AI technology is not only a need for career development but also an opportunity to participate in shaping the future.

However, the learning curve for generative AI is steep: the field is developing rapidly, with new technologies, frameworks, and applications emerging endlessly. Learners need to cross multiple knowledge domains from Transformer architecture to Agent systems. The Gen-AI-Learning project, as an exploration record of learners, provides references and inspiration for peers.

The learning path of Gen-AI-Learning revolves around four core domains:

1. Natural Language Processing (NLP)：As the foundation of generative AI, covering word vectors, sequence models, attention mechanisms, and Transformer architecture;
2. Large Language Models (LLM)：Focus on learning the architecture, training methods, inference optimization, and practical skills like prompt engineering for models such as GPT series, Llama, and Mistral;
3. LangChain：A bridge connecting LLMs and applications, providing core abstractions like chain calls, tool integration, and memory management;
4. LangGraph：The latest evolution of LLM application architecture, introducing graph structures to support complex control flows and state management.

Gen-AI-Learning emphasizes practice orientation:

• Hands-on experiments: Implement attention mechanisms from scratch, build simplified GPT models to understand architectural principles;
• Project practice: Build small applications like text summarization and question-answering systems early to comprehensively apply the learned knowledge;
• Errors and debugging: Record confusion, detours, and bug-fixing processes—these 'failure experiences' have important teaching value.

LangChain is a core framework for LLM application development:

• Chain: Encapsulates steps like LLM calls and prompt templates into composable components;
• Tool integration: Expands the boundary of LLM capabilities, supporting external interactions such as search engines and database queries;
• Memory: Solves the context window limitation, providing multiple memory implementations;
• Retrieval-Augmented Generation (RAG): Combines external knowledge bases to alleviate model knowledge cutoff and hallucination issues.

LangGraph addresses the limitations of LangChain's linear model by introducing graph structures:

• Graph structure: Supports complex control flows like loops, branches, and parallelism;
• State management: Provides type-safe state definition and persistence;
• Agent workflow: Supports complex autonomous systems like ReAct mode and multi-agent collaboration;
• The project records the learning transition from LangChain to LangGraph and migration examples.

Gen-AI-Learning organizes and contributes open-source learning resources:

• Paper reading: Records key points and insights from important papers (e.g., Attention Is All You Need);
• Open-source projects: Encourages reading source code of LangChain, LlamaIndex, etc., to understand industrial-level engineering practices;
• Community participation: Shares experiences of participating in open-source communities, technical forums, and academic conferences to expand horizons and networks.

Learning insights:

• Balance breadth and depth: First establish a global perspective, then dive into interested directions;
• Combine theory and practice: Use theory to guide practice, and practice to deepen theory;
• Continuous learning: Maintain curiosity and track the latest developments in the field;
• Value of sharing: Recording notes helps others while deepening one's own understanding.

Outlook: The project will continue to update to track the latest progress in generative AI; it encourages learners to take action—through systematic learning, practice, and community participation, find their own place in this field.