Join me for a deep dive into the world of Large Language Model (LLM) multi-agent systems. In this session, we will look into how to orchestrate multiple AI “agents” to collaborate, share tasks, and produce higher-quality results than a single LLM could on its own.
You’ll discover:
- Core Concepts: Why multi-agent ecosystems matter, what “agent autonomy” and “collaboration” look like in practice, and how planning and reflection enhance outcomes.
- Live Demos: Step-by-step walkthroughs of Python code using open-source frameworks like AutoGen, showcasing how agents can be equipped with specialized tools (e.g., web scraping, code execution) and then work together.
- Design Patterns: From simple question-answer agents to complex multi-agent collaboration, see how to set up reflection loops, tool usage, and role-based planning.
- Practical Applications: Learn how these workflows can power use cases in support systems, data analysis, report writing, and more—complete with important guardrails and best practices.
- Challenges & Guardrails: Understand the potential pitfalls of large-scale agent systems, including cost, complexity, and the steps you can take to mitigate risk (sandboxing code execution, limiting conversational turns, etc.).
If you’re looking to build “gigantic” LLM-based workflows—whether to automate tasks, handle complex data analysis, or integrate advanced AI capabilities into your existing apps—this talk is for you.
Enjoy the session, and be sure to explore the sample code and links I shared during the talk to kick-start your own multi-agent AI projects!
00:00 – Introduction & Speaker Bio
A brief welcome to the talk and formal introduction of our speaker, Daron Gundam.
00:33 – Daron’s Background & Session Overview
Highlights of Daron’s professional experience and a quick look at what this session will cover.
01:00 – Why LLM Systems Matter
An explanation of the importance of Large Language Model (LLM) workflows and the value of building multi-agent systems.
06:04 – The Three Foundations (Autonomy, Collaboration, Adaptability)
A deep dive into the core principles behind effective multi-agent LLM systems.
09:49 – Key Design Patterns (Reflection, Tool Use, Planning)
Overview of the main patterns that guide multi-agent interactions and improve output quality.
14:00 – Multi-Agent Collaboration Overview
How different agents, each with specific roles, can work together to solve complex tasks.
15:00 – Demo 1: Single-Agent Workflow
A hands-on example showing how to set up and execute a simple single-agent LLM process.
23:36 – Demo 2: Multi-Agent Web Scraping
Showcasing an LLM agent equipped with web-scraping tools to autonomously gather information.
36:29 – Advanced Workflows & Tools Integration
Strategies for integrating diverse tools and managing more sophisticated, large-scale LLM tasks.
40:31 – Demo 3: Manager, Data Gatherer, Analyst, Writer
How multiple specialized agents can coordinate to produce a comprehensive report.
43:40 – Real-World Use Cases & Challenges
Practical applications, potential pitfalls, and considerations when implementing multi-agent systems.
46:20 – Q&A Session
Audience questions about sandboxing, code execution, logging, and best practices for agent orchestration.
52:06 – Closing Remarks & Wrap-Up
Final thoughts on multi-agent systems, key takeaways, and pointers to additional resources.