Hyogon Ryu
LLM Engineering Team Lead at KRAFTON AI
I am a research engineer and lead the LLM Engineering Team at KRAFTON AI, working with Kangwook Lee.
I received my M.S. from KAIST, advised by Hyunjung Shim, and my B.S. in Cyber Defense from Korea University. My research was on diffusion model quantization; before that, I was a reverse engineer on CyKor, the best white-hat hacking team in South Korea.
Research Focus
I work on ML systems optimization & engineering across pretraining, RL, and inference. We’re a small team, so we work across the stack and tackle whatever needs optimizing.
Work
- 2026Raon. KRAFTON’s foundation model family. Hugging Face
- 2025SmartZoi. On-device, real-time social AI for the life simulation game, inZOI. NVIDIA video
Publications
Online Agent-as-a-Judge: Situation-Generating Evaluation for Interactive Agents
Terminus-KIRA: Boosting Frontier Model Performance on Terminal-Bench with Minimal Harness
DGQ: Distribution-Aware Group Quantization for Text-to-Image Diffusion Models
Memory-Efficient Fine-Tuning for Quantized Diffusion Model
ECCV 2024Code
Experience
- LLM Engineering Team Lead, AI Model Research Dept., KRAFTON AI
- Research Engineer, Simulation AI Team, KRAFTON AI
- Software Engineer Intern, NAVER
Education
- M.S. in Artificial Intelligence, KAIST
- B.S. in Cyber Defense, Korea University
- Korea Science Academy of KAIST