Undergraduate researcher, Electrical Engineering, Korea University
I study statistical and mathematical guarantees for learned models in robotics: when a robot can trust a safety constraint that rests on learned perception or a learned world model.
This fall I am an exchange researcher in the Machines in Motion Laboratory at NYU Tandon with Prof. Ludovic Righetti, studying how physically grounded conformal safety guarantees carry into a learned world model and the control modules downstream of it, and how far they hold under distribution shift. At Korea University I work in the HPIC Lab with Prof. Jong-Kook Kim.
I am applying to PhD programs for Fall 2027.
NeurIPS 2026 Workshop on Robot Learning with World Models, poster
When planning moves into a learned latent space, the statistical coverage of a conformal safety margin survives, but its meaning in metres and its usable geometry do not. A decoder head that reads clearance back in metres restores both.
arXiv:2610.37217, 2026. Journal version in preparation.
A directional, time-varying conformal keepout whose margin is selected per viewing direction from the sensing severity, enforced at plan time and at every command without relaxing its coverage level.
Under review
An attention-based mapper that assigns circuit qubits to hardware qubits for scalable quantum compilation.