Theory and Construction
of Agent Systems AI Agent Engineering 101

What you learn here does not move when AI trends do.

INSTRUCTOR

Introduction

Jaegyu Lee lecturing
Jaegyu Lee 이재규
SOFTWARE ENGINEER / OPEN SOURCE MAINTAINER

I am a software engineer of ten years, currently at ZEP. I finished my master's at Seoul National University of Science and Technology, where this course is taught, researching data pipelines and distributed systems, with papers at BIGCOMP and in Frontiers in Big Data.

At ZEP I run a WebRTC cluster serving 4M MAU. I moved it from LiveKit Cloud to self-run OCI infrastructure and cut infrastructure cost by 60%.

I build Ouroboros, an open-source Agent OS: a framework that pins requirements down into a Seed through interviews and verifies execution results with deterministic gates. ZEP QUIZ (over 600K quizzes a month) runs an agent that improves its own tools. I presented this architecture at the ICML 2026 AI×Education workshop.

LECTURES
  • Seoul National University, College of Innovation: design thinking and Socratic reasoning in the Silicon Valley AI era 2026
  • Seoul National University, College of Innovation: AI education 2026
  • Seoul National University Silicon Valley tour: AI Native 2026
  • KT AI capability program, Harness Engineering: controlling probabilistic AI as a deterministic system 2026
  • Hanwha AX lecture 2026
  • Oracle Digital Native Roundtable: building an RTC cluster for 4M MAU 2026
COMMUNITY
  • Ralphthon #1 winner, Ralphthon #2 judge and speaker
  • GDG Korea-Japan hackathon judge
  • GDG Build with AI Incheon, session on the CS fundamentals harness engineering needs
  • Hermes Meetup RLM-FORGE session, Arize AI meetup Agent OS session
SYLLABUS

15 Weeks

The first eight weeks are the academic track. We establish the definitions and the genealogy, then put classical MAS and modern protocols through a reproduction experiment each week. The last seven weeks are the project track. Teams build an agent transaction stack (discovery, negotiation, execution, payment, ledger) and cross-test it on demo day.

PHASE 1: DEFINITIONS, GENEALOGY, PROTOCOLS
PHASE 2: BUILDING THE AGENT TRANSACTION STACK
  • W09Project Kickoffteam formation
  • W10Mini-App Agent Designdesign
  • W11Negotiation Between Agentsnegotiation
  • W12Memory and Selective Disclosurememory / disclosure
  • W13Payment, Settlement, Ledgerpayment / ledger
  • W14Cross Testingcross testing
  • W15Demo Daydemo day