Machine Learning Engineer since 2021.
Focus: 3D Deep Learning, Generative Models, and Real-time Inference.
Currently building 3D content generation systems using large diffusion and language models.
Main interest: making ML models run faster, cheaper, and more useful in production (especially for graphics / game pipelines).
- Single-image-to-3D generation on RTX 2060 (<15s runtime)
β Deployed in inZOI, used by players to generate 3D furniture and props from photos
β Widely shared by users across platforms: - Real-time text-to-building system using CLIP retrieval
- Fast PBR texture inpainting using diffusion + ControlNet
- ACM TOG (proc. of SIGGRAPH) paper, US patent
- Languages: Python, C++, JavaScript
- Frameworks: PyTorch, TensorFlow, ONNX
- Graphics: CUDA, OpenGL, WebGL
- Infra/Tools: Unreal Engine, Blender
- Working at KRAFTON, developing 3D generation features (e.g. in inZOI)
- Prev: ML engineer at SPACEWALK, focused on architecture optimization using RL
- MS in CS @ KAIST (Thesis on Monte Carlo rendering denoising)
- BS in CS and Math @ KAIST (Summa cum laude)
- π§ ciy405x@gmail.com
- π Seoul, South Korea
- π LinkedIn