Hyogon Ryu

Applied Research Scientist, Krafton AI

Email / CV / Google Scholar / GitHub / LinkedIn

I am an Applied Research Scientist at Krafton AI, where I work on simulation and agent systems. My current interests are agentic AI, social simulation, and efficient AI systems.

I completed my M.S. in AI at KAIST in February 2025, advised by Prof. Hyunjung Shim, and earned my B.S. in Cyber Defense from Korea University. Before working on agents and simulation, I was active in white-hat hacking and CTFs as a member and vice president of CyKor.

Research Interests

• Coding agents and agentic workflows. How agents plan, execute, verify, and communicate progress inside real software projects.
• Social simulation with AI. How believable characters, memory, and interaction loops can make virtual worlds feel alive.
• Efficient AI systems. Quantization and memory-efficient adaptation for generative models.
• AI and people. How frontier technology changes the behavior, taste, and happiness of people who live with it.

Research and Professional Work

• SmartZoi: character simulation work for inZOI at Krafton AI, focused on making game characters behave more believably through AI-driven simulation. Video
• Terminus-KIRA: coding-agent harness built at Krafton AI by a task force in about one month. It reached 74.8% on Terminal-Bench 2.0, ranked #2 globally and #1 with Opus 4.6. Project / Tech Blog
• DGQ: Distribution-Aware Group Quantization for Text-to-Image Diffusion Models
  Hyogon Ryu, NaHyeon Park, Hyunjung Shim. ICLR 2025. Paper / Code
• Memory-Efficient Fine-Tuning for Quantized Diffusion Model
  Hyogon Ryu, Seohyun Lim, Hyunjung Shim. ECCV 2024. Paper / Code

Side Projects

• Playground / ai-npc-world: a small open-world village on this homepage whose AI NPCs remember visitors. ai-npc-world is the lightweight package split out while building Playground. Playground / GitHub

Experience

• Krafton AI (2025.04-Present)
  Research Scientist, Simulation AI Team. Working on inZOI character simulation and SmartZoi.
• Naver Corp. (2022.01-2022.02)
  Software Engineer Intern, Multimodal Applications. Worked on trivia sentence extraction from unstructured text data and search engine improvement.
• AI Research Lab, Korea University (2021.01-2021.12)
  Undergraduate Research Student. Worked on energy backdoor attacks on deep learning models.

Education

• M.S. in Artificial Intelligence, KAIST, 2023.02-2025.02
  Advisor: Prof. Hyunjung Shim. Focus: efficient deep learning systems and diffusion model quantization.
• B.S. in Cyber Defense, Korea University, 2019.03-2023.02
  Interests: reverse engineering, system hacking, and AI for security.

Honors

• DEF CON CTF Final 2020: #8, Team CyKor
• CODEGATE CTF Univ. Final 2020: #6, Team CyBie
• SECCON CTF 2022: #4, Team Korean
• CLOVA AI RUSH 2021: #2 in Round 1 and #4 in Round 2
• Cyber Security AI Bigdata Challenge 2021: #2 in Final, Team Leader