LangGraph and Agentic AI Syllabus

1. Header Information

Field	Value
Technical Group	AI Engineering
Topic Name	LangGraph and Agentic AI
Topic Code
Version	1.0
Training Audience	Freshers / Interns with basic programming knowledge (Python)

2. Course Objectives

This topic deep dives into building advanced agentic AI systems using the LangGraph framework, moving beyond linear chains to cyclic, stateful workflows. Trainees will master core LangGraph concepts such as State Management, Nodes, and Edges, while implementing advanced patterns like ReAct, Planning, and Multi-Expert orchestration. The training covers the integration of specialized tools with Tavily Search, multi-agent collaboration strategies (hierarchical), and Human-in-the-Loop mechanisms with persistent state checkpointing to build production-ready agentic applications.

[!NOTE] This topic equips trainees with the skills to architect complex multi-agent systems, adapting them to real-world enterprise scenarios that require persistence, human oversight, and specialized expertise.

Output Standards table

Name	Code	Description
LangGraph Core		Master LangGraph architecture, state management, and message-centric patterns.
Agentic Patterns		Understand and implement ReAct, planning, reflection, and multi-expert patterns.
Tool Integration		Integrate external tools and search APIs efficiently within agent workflows.
Multi-Agent System		Design and coordinate multi-agent teams and hierarchical systems.
Human-in-the-Loop		Implement persistence, memory, and human approval mechanisms.

Detailed Learning Outcomes (LOs)

• LangGraph Foundations

  • LO-1 - Understand LangGraph architecture and the role of Messages in State.
  • LO-2 - Master State Management with messages-centric patterns and distinguish between messages (I/O) and context (metadata).
  • LO-3 - Build cyclic workflows using Nodes and Edges with LangChain integration.

• Agentic Patterns

  • LO-4 - Implement Research Agent using the ReAct (Reasoning + Acting) pattern.
  • LO-5 - Use LangGraph's prebuilt ToolNode and apply advanced techniques like reflection and planning.
  • LO-6 - Apply best practices for designing production-ready autonomous agents.

• Tool Calling & Integration

  • LO-7 - Understand Tool/Function Calling mechanisms in LLMs.
  • LO-8 - Integrate external tools and use Tavily Search API for optimized web searching.
  • LO-9 - Manage tool orchestration and handle parallel tool execution.

• Multi-Agent Collaboration

  • LO-10 - Design and implement multi-agent architectures (Sequential, Hierarchical).
  • LO-11 - Coordinate agent workflows, master context injection, and managed state transitions.
  • LO-12 - Implement logic for agents to decide when to end, escalate, or hand off tasks.

• Human-in-the-Loop & Persistence

  • LO-13 - Implement human approval workflows and breakpoints in agent execution.
  • LO-14 - Persist agent state using checkpointers and implement memory management.
  • LO-15 - Apply time-travel and state editing for debugging and reliability.

3. Topic Outline

• Unit 01: LangGraph Foundations & State Management
• Unit 02: Agentic Patterns: Multi-Expert Research Agent
• Unit 03: Tool Calling & Tavily Search
• Unit 04: Multi-Agent Collaboration
• Unit 05: Human-in-the-Loop & Persistence

4. Time Allocation

Activity Type	Percentage	Description
Concept/Lecture	40%	Concepts, theory (LangGraph Core, Agentic Patterns, Multi-Agent Collaboration)
Assignment/Lab	40%	Hands-on Labs (ReAct Agent implementation, Multi-Agent orchestration, HITL workflows)
Guides/Review	10%	Code reviews, Q&A, debugging state and persistence
Test/Quiz/Exam	10%	Daily Quizzes, Final Project (Multi-Agent System)

5. Training Materials & Environments

Textbooks & Guides

• Internal Knowledge Base:
  • LangGraph Foundations & State Management
  • Agentic Patterns: Multi-Expert Research Agent
  • Tool Calling & Tavily Search
  • Multi-Agent Collaboration
  • Human-in-the-Loop & Persistence

References

• LangGraph Documentation
• LangChain Documentation

Technical Requirements

• Python 3.10+
• Jupyter Notebook / VS Code
• OpenAI API Key (or equivalent LLM provider)
• Tavily API Key
• Libraries: langgraph, langchain, langchain_community.tools.tavily_search

6. Assessment Scheme

Component	Quantity	Weight (%)	Note
Quiz	?	?%	Daily quizzes after each unit
Assignments	?	?%	Lab exercises
Final Project	?	?%	Final project implementing a multi-agent system

Pass Criteria

• Total topic GPA >= 60/100
• Completed 100% Assignments and Final Project

7. Training Delivery Principles

Role	Responsibility/Criteria
Trainees	Passed previous basic AI modules. Active participation in labs.
Trainer	Senior AI Engineer with hands-on LangGraph experience.
Training	Daily sessions followed by practical coding labs.
Re-Test	Allowed once for the Final Project if GPA < 70 but >= 50.
Marking	Code quality, functionality, and understanding of concepts.