As employees in the 21st century, we face an inevitable reality: artificial intelligence is transforming the workplace faster than ever before. I don’t believe it’s a matter of if our jobs will be affected — it’s a matter of when. In China, AI-powered robots now assemble cars in “dark factories,” where production continues around the clock without the need for human supervision — or even lights. Intelligent AI agents aren’t just coming; they’re already here.
As a cybersecurity analyst, I recognize that my field is no exception. Many of the tasks we perform today can and likely will be automated by intelligent agents. That’s why I believe it’s essential for professionals like us to understand how these systems work — not to fear them, but to learn how to design, maintain, and secure them. If we want to stay relevant in the era of autonomous agents, we need to be the ones building and managing them.
This project marks my first step in that direction. Through a series of experiments and workflows, I aim to build hands-on familiarity with LangGraph and LangChain, exploring how agentic systems can be designed, orchestrated, and applied to real-world problems.
A special thanks to Mr. Hacker Loi.
According to LangChain's official documentation:
Workflows have predetermined code paths and are designed to operate in a certain order. Agents are dynamic and define their own processes and tool usage.
This repository is part of my ongoing series of projects designed to build familiarity with LangGraph and LangChain—two powerful frameworks for developing AI-driven systems that combine reasoning, memory, and structured task orchestration.
Each part of this series focuses on a specific aspect of intelligent agent design, starting from static, rule-based workflows and progressing toward dynamic, autonomous AI agents.
- Understand how LangGraph structures logic using states, nodes, and edges
- Learn to design workflows that control and monitor AI behavior predictably
- Gain practical experience integrating LLMs, retrieval tools, and self-evaluation loops
- Transition from workflows → multi-agent systems capable of decision-making and collaboration
| Part | Title | Description | Link |
|---|---|---|---|
| 1 | Introduction to Workflows with LangGraph | Implements a LangGraph-based IT support ticket system using a structured workflow. The system classifies support tickets, retrieves knowledge base entries, drafts responses, and self-evaluates them. | Part 1: Workflows |
| 2 | Enhancing Workflows (coming soon) | Extends Part 1 by fixing draft generation issues and adding improvements to workflow efficiency, self-evaluation, and output quality. This part builds directly on the previous system, laying the groundwork for autonomous agents. | In progress |
| 3 | Building AI Agents with Bedrock Agent Core (coming soon) | Introduces an autonomous agent built on Bedrock Agent Core that analyzes WAF logs and provides actionable security insights. This part demonstrates real-world applications of agentic reasoning and decision-making in cybersecurity. | In progress |