I am a postdoctoral researcher at POSTECH Institute of AI (PIAI), working with Prof. Minsu Cho, who also advised my Ph.D. in Artificial Intelligence at POSTECH GSAI.

I work on 3D visual perception for systems that act in the physical world. Attaching vision to a body changes what perception has to do. It needs open-vocabulary 3D segmentation (Mosaic3D) fast enough for a control loop, and active perception (Affostruction) to decide where to look next when one view is not enough. Both draw on machinery I am still developing, efficient 3D architectures (Fast Point Transformer, PointMixer) and geometric deep learning for pose-independent shape representation (CHOIR, RIST). I am building 3D world action models for robot manipulation, where a model perceives, predicts, and acts in one 3D space.