Postdoctoral researcher at POSTECH Institute of AI, working with Prof. Minsu Cho.

I work on 3D visual perception for systems that act in the physical world, drawing on efficient 3D architectures (Fast Point Transformer, PointMixer) and geometric deep learning (CHOIR, RIST). Acting demands open-vocabulary 3D perception at control-loop speed (Mosaic3D) and active perception when one view falls short (Affostruction). Both meet in the 3D world action models I build for robot manipulation, which perceive, predict, and act in one 3D space.