Now

Work

I am a Compiler Engineer in the Compiler Team at Hyundai Motor Company. My work involves TVM-based compilation for custom NPUs, including lowering, scheduling, and memory mapping.

I completed my Ph.D. at Yonsei University in August 2026, advised by Prof. Dosik Hwang.

Research

Test-time adaptation remains my main research interest. I am interested in what a model can safely learn from incoming, unlabeled data, and how that affects its predictions under distribution shift.

Medical image segmentation and reconstruction are recurring applications. Efficient deployment is another part of the same question: how to make a model useful when its data and execution environment differ from the conditions in which it was trained.

Reading & building

• Adaptation mechanisms that change a small part of a pretrained model.
• Tensor programs, memory reuse, and the path from Relax to TIR.
• Small experiments that make these mechanisms easier to inspect.

Some of these ideas are collected in the research notes and interactive lab.

Get in touch

For research and professional enquiries, email is the best way to reach me. My CV has the longer record.