Learning from Imperfect Data

I work on label-efficient visual learning. My research asks how visual models can learn useful structure when data is costly to annotate, difficult to store, or imperfect in practice. I explore limited supervision and compact representations, spanning weakly supervised and language-guided segmentation, object localization, and storage-efficient training.

I received my M.S. and Ph.D. in Integrated Technology from Yonsei University, advised by Prof. Hyunjung Shim. During my research internship at NAVER AI Lab, I worked with Song Park, Byeongho Heo, and Dongyoon Han on storage-efficient visual training.

I am currently an AI researcher and Staff Engineer at Samsung Electronics' AX/PI Center, working on semiconductor data modeling and layout generation. These industrial settings bring the same research question into practice: learning from structured patterns and noisy data with limited supervision.