生成式AI实战课：从LangChain到RAG与模型微调_CG群

本课程专为有Python基础的开发者设计，系统讲授LangChain、向量RAG、GraphRAG及MCP工具集成，并涵盖LoRA/QLoRA大模型微调与部署，助你从API调用进阶到企业级AI应用开发。

原始标题：Generative AI: RAG, LangChain, GraphRAG & Fine-Tuning

Generative AI: RAG, LangChain, GraphRAG & Fine-Tuning

Deesa Technologies 推出的生成式 AI 实战开发课程专为具备 Python 基础的开发者设计，旨在通过构建基于 OpenAI/DeepSeek API 的 Gradio 应用，系统讲授 LangChain、向量 RAG、GraphRAG、MCP 工具集成以及智能体构建等核心技术。该课程不仅涵盖前沿技术，还包含 LoRA/QLoRA 大模型微调及部署，助力学员实现从 API 调用到企业级 AI 应用开发Pipeline的跨越。

Build LLM apps with MCP tools, guardrails, memory & Hugging Face deploy

What you’ll learn
⚡ Build your first Generative AI applications using LLM APIs (OpenAI, DeepSeek, Groq) and Gradio UIs
⚡ Understand how LLMs work — tokens, probability, capabilities, limitations, cost, and latency
⚡ Design effective prompts and tune model parameters like temperature for consistent outputs
⚡ Evaluate AI model quality using scorecards and LLM-as-judge patterns
⚡ Apply prompt engineering techniques and implement guardrails for safer, more reliable AI apps
⚡ Extract structured data from documents (e.g. contract clauses) using LLM-powered pipelines
⚡ Build LangChain applications with tool integration, web search, and LCEL chains
⚡ Create an AI SQL Assistant that generates, validates, and runs database queries safely
⚡ Compare stateless LLMs vs LangChain memory strategies for multi-turn conversations
⚡ Build Retrieval-Augmented Generation (RAG) systems with vector stores over real documents
⚡ Use GraphRAG to connect knowledge graphs with vector search for relationship-aware AI
⚡ Build MCP-powered AI agents that use tools to query data, documents, and external systems
⚡ Summarize long documents using map-reduce pipelines when content exceeds context limits
⚡ Fine-tune open-weight models using LoRA and QLoRA on real customer support examples
⚡ Deploy a fine-tuned model to Hugging Face Spaces with a production-style Gradio demo
⚡ Structure Gen AI projects the way engineering teams do — config, prompts, services, and guardrails

Requirements
❗ Basic python programming is preferred

DescriptionGenerative AI: RAG, LangChain, GraphRAG & Fine-Tuning is a hands-on course fromDeesa Technologies for developers who want tobuild real Generative AI applications — not just call an API once and stop.

You’ll follow a clear path from yourfirst LLM app throughproduction-style patterns used in modern engineering teams: prompt design, guardrails, structured outputs, LangChain, conversation memory, RAG, GraphRAG, MCP tool integration, map-reduce summarization, andfine-tuning open-weight models with deployment toHugging Face.

Every module includesrunnable projects with a consistent structure (main, services, config, prompts, guardrails) so you learn how professionals organize Gen AI codebases.

What you’ll build
Across11 modules, you’ll create practical applications including

✨Executive Brief Generator andProfessional Report Drafter — your first Gradio LLM apps

✨Prompt Playground andTemperature & Latency Lab — control model behavior, cost, and speed

✨Model Scorecard andTranslation Quality Report — evaluate outputs with real evaluation patterns

✨Output Formatter andContract Clause Extractor — structured extraction from documents

✨AI SQL Assistant — LangChain, LCEL, and safe natural-language-to-SQL

✨Customer Support Thread Analyzer — stateless vs memory-based conversations

✨On-Call Runbook Assistant — RAG over production runbooks

✨GraphRAG Dependency Explorer — combine graphs + retrieval for relationship-aware answers

✨MCP-powered assistants — connect LLMs to databases, knowledge bases, and tools via Model Context Protocol

✨Long Document Brief Generator — map-reduce summarization for books and large files

✨Customer Support AI — QLoRA fine-tuning on open-weight models + Hugging Face Spaces deployment

Who this course is for
⭐ Developers and engineers who want to move from “calling an API” to building real Gen AI applications
⭐ Python learners comfortable with basic coding who want hands-on AI project experience
⭐ Software engineers upskilling into AI — backend, data, or full-stack developers adding Gen AI to their toolkit
⭐ Tech professionals who need practical patterns: prompts, guardrails, RAG, agents, and deployment — not just theory
⭐ Builders who learn by doing — every module includes runnable projects with production-style code structure
⭐ Anyone preparing for real-world AI work — support bots, document Q&A, SQL assistants, runbook tools, fine-tuned models
⭐ Students and career switchers with Python basics who want a clear path from first LLM app to Hugging Face deployment

隐藏内容

此处内容需要权限查看