Enterprise AI Agents Development Bootcamp
From zero to one, build 3 complete Agent projects based on LangChain/LangGraph, covering tool-based, conversational, and multimodal paradigms. Master production-grade AI Agent full-cycle development and deployment.
Course Overview
Lack of Agent Development Methodology
LLM capabilities evolve rapidly, but there is no systematic development process from research and project initiation to product design and engineering delivery, resulting in inconsistent project quality.
Rapid Framework Iterations Hard to Keep Up
Frameworks like LangChain and LangGraph update frequently with breaking changes. Fragmented documentation makes it difficult to build a complete knowledge system, leading to high learning costs.
Lack of Real-World Project Experience
Most tutorials stop at demos, lacking the complete hands-on path from MVP to production-grade release, unable to handle complex requirements in real business scenarios.
Gap in Production Deployment Skills
Model development and engineering deployment are severely disconnected, lacking critical skills such as unit testing, containerized deployment, and service publishing to bring Agents into production.
Course Projects
All course project code is open-sourced on GitHub, with a cumulative 900+ Stars. Fork and build upon them after completing the course to continuously iterate on your own Agent products.
GitHubSentinel
364Intelligent information retrieval Agent that automatically tracks GitHub open-source projects and Hacker News trending topics, generating daily/weekly reports.
LanguageMentor
74Online language tutor Agent powered by LLaMA 3, supporting 100+ languages with scenario-based conversations and vocabulary building.
ChatPPT
174Multimodal intelligent assistant supporting voice, image, and text input to automatically generate PowerPoint presentations.
Core Technology Stack
LangChain / LCEL
Build complex applications like RAG and Multi-Chain using LangChain v0.2 latest APIs and LCEL expression language, with LangSmith for end-to-end debugging.
LangGraph
Master graph-based orchestration (nodes/edges/state) to implement multi-turn conversations, tool calling, Multi-Agent collaboration, and reflection mechanisms for production-grade Agent architectures.
Ollama
Private LLM management tool supporting local deployment and service integration of open-source models like LLaMA 3 and GLM-4, enabling data-sovereign security solutions.
Gradio
AI-focused frontend framework, from basics to Gradio 5 advanced features. Quickly build interactive GUI interfaces with support for voice, image, and other multimodal inputs.
Core Value
3 Complete Agent Projects
The entire course is driven by three real-world projects: GitHubSentinel (tool-based), LanguageMentor (conversational), and ChatPPT (multimodal), each iterated from v0.1 to production-grade release.
Full-Cycle Product & Engineering Coverage
Master the complete pipeline in one course: market research, product design, prompt engineering, Agent development, frontend integration, unit testing, and Docker containerized deployment.
ChatGPT Pair Programming Methodology
Learn best practices for pair development with ChatGPT. Use multi-turn prompts to generate project-level code from scratch and iterate continuously, dramatically boosting development efficiency.
Cutting-Edge LLM Hands-On Implementation
Build productivity-boosting AI Agents for real work scenarios using cutting-edge models including GPT-4, LLaMA 3, Whisper Large v3, and MiniCPM.
Learning Outcomes
- Master Agent development methodology and independently complete the full process from research to production deployment
- Proficiently use LangChain v0.2 / LCEL / LangGraph to build single-agent and multi-agent systems
- Gain hands-on development capabilities across three major Agent paradigms: tool-based, conversational, and multimodal
- Master production-grade engineering skills including Ollama private deployment and Docker containerized publishing
- Apply LLM capabilities to real business scenarios such as information subscription, language tutoring, and office automation
Detailed Syllabus
01 Tool-Based Agent — GitHubSentinel
Agent Development Methodology & Project Initiation
- Agent overview, core tech stack, and typical development paradigms
- Best practices for pair development with ChatGPT
- GitHub Sentinel market research and MVP feature design
- Using GPT-4 multi-turn prompts to generate initial version code
Analysis Reports & Conditional Filtering
- GitHubClient implementation for project subscription and progress retrieval
- OpenAI GPT service integration and automated report generation
- Report generation with status and period-based filtering
- Prompt exploration optimization and log management
Gradio Frontend & User Interface Design
- Gradio fundamentals and best practices
- Gradio GUI design and integration
- Command-line interaction and background scheduled task mode
- Prompt optimization based on System Role
Scheduled Updates, Email Notifications & Private Deployment
- Python daemon processes and schedule library integration
- SMTP email notification feature design and implementation
- Ollama private LLM management and deployment
- GitHubSentinel integration with private LLM services
Advanced Features & Production-Grade Release
- unittest unit testing and Docker containerized deployment
- Hacker News trending topic report extension
- Gradio advanced: adapting to multiple report types and LLM services
- Multi-task scheduling and fully-featured production-grade Agent
02 LangGraph Deep Dive
LangChain New Ecosystem & LCEL Primer
- LangChain evolution review and breaking changes in new versions
- LangSmith / LangGraph / LangServe tech stack walkthrough
- LCEL core: Runnable protocol and Multi-Chain applications
- Hands-on: Building a RAG application with LCEL
Building Production-Grade AI Agents with LangGraph
- LangGraph design philosophy: graphs, nodes, edges, and state
- Checkpoints (memory) and interrupts (human-in-the-loop)
- LangGraph tool integration and LangSmith debugging
- Hands-on: Multi-turn conversational customer service (chatbot)
LangGraph Multi-Agent Systems
- Tool Calling and custom Agent nodes
- AgentNode templates and inter-agent communication via AgentState
- ToolNode prebuilt components and multi-agent collaboration
- Conditional edges, routing functions, and Multi-Agent hands-on practice
LangGraph Reflection Mechanism Agents
- Three key elements of Agent construction: Prompt / LLM / Tools
- Agent persona design and workflow orchestration
- Reflection & Critique mechanism principles and hands-on practice
- LangGraph technical summary and advanced considerations
03 Conversational Agent — LanguageMentor
Market Research & Product Design
- Online language learning market size, growth trends, and pricing model analysis
- LanguageMentor product design: conversation training and scenario-based experiences
- Core technical approach: ChatBot + Contextual Prompt
- Agent application tech stack: LangChain + Ollama + Gradio
Scenario Conversation Feature Design & Development
- In-depth analysis of Hi Echo conversation and scenario features
- Scenario-based experiences: hotel check-in, job interviews, and more
- Guided conversations: from open-ended chat to directed dialogue
- LangChain Runnables message history and Gradio responsive layout
Vocabulary Building & Advanced Prompting
- Structured prompt design, dynamic conditional logic, and multi-turn conversation techniques
- Gamified level-based vocabulary building feature design
- Rapidly extending personalized Agents based on AgentBase
- Gradio Tab components for decoupling different feature modules
04 Multimodal Agent — ChatPPT
Market Research & Product Design
- Enterprise office scenario pain points and automation market analysis
- PowerPoint content generation approach and style template design
- ChatPPT input content format and layout management
- ChatPPT v0.1 prototype design and implementation
Text-to-Content Generation Development
- python-pptx core data structures and operations
- Custom data structures and layout strategy with LayoutManager
- Extensible single-slide generator with SlideBuilder
- Designing prompt engineering for PowerPoint content generation
Topic Generation & Gradio 5 Integration
- Gradio 5 new features overview and AI-generated code
- Prompt design for topic-to-PowerPoint content generation
- Integrating LangChain ChatBot with memory capabilities
- Encoding-based layout group strategy and end-to-end generation
Voice Input & Domain Publishing
- OpenAI Whisper Large v3 ASR service development
- Speech recognition integration and audio file processing
- Hierarchical content layout: multi-level BulletPoints and minimalist master slides
- Domain configuration, SSL certificate issuance, and Nginx reverse proxy
Image Recognition & Multimodal Input
- Multimodal model MiniCPM-V integration and image recognition
- Gradio integration of image recognition features
- Word document parsing and image embedding for multimodal input
- Intelligent image suggestion Agent (Image Advisor) design and implementation
Smart Image Matching & Containerized Release
- LLM-based topic content search and image matching technology
- Search engine selection and cost model analysis
- Using ChatGPT to generate unit test code
- Docker image building and ChatPPT containerized deployment
Instructor
Jingtian Peng
Founder / CEOFounding member of Huawei 2012 Lab deep learning team. UC visiting scholar. Former technical partner at Caicloud (acquired by ByteDance in 2020), co-founder & CTO of Pinlan Data (raised ~¥200M). Kubeflow maintainer, TensorFlow contributor, Linux CNCF program committee member.
