Computer vision and generative AI engineer, based in Korea.
For the past two years I have been building image and video generation pipelines that run in production: diffusion models orchestrated with ComfyUI, served through FastAPI and Docker on GPU servers, with offscreen 3D rendering to produce depth, normal and part maps for conditioning. Much of my recent work is about keeping these pipelines fast and within GPU memory: deciding which models stay loaded, which ones are released between steps, and sharing models across pipelines.
ComfyUI-SMPL: ComfyUI nodes that turn a height and a weight into an SMPL / SMPL-H body, then output a shaded render, an exact silhouette mask, a depth map and 24 body-part masks for ControlNet-style conditioning.
Rendered from the SMPL / SMPL+H models, © Max-Planck-Gesellschaft, shown for illustration under their non-commercial license.
- Flicker: flicker reduction in old films with Midway equalization (MATLAB, team of three, 2019).
- Tic-tac-toe: a tic-tac-toe AI with minimax (OCaml, 2018).
- Advent of Code: my yearly puzzle solutions in Python.
- Master's in Computer Science, Image and AI track, Université de Bourgogne, France (2020)
- Research internship at INRAE Dijon: measuring pea root architecture from segmented images (2020)
- Thermal camera calibration study at the ImViA lab (2019–2020)
- Coding instructor for children at Code Camp, Sydney (2023–2024)
Python · PyTorch · ComfyUI · OpenCV · Blender (bpy) · pyrender · FastAPI · Docker · ONNX · Git · C++ (secondary)
French (native) · English (professional) · Korean (intermediate)