Hello!

I am a sophomore at the University of Southern California, studying Applied and Computational Mathematics with a Minor in AI Applications. I am currently working as a Research Assistant in the LiraLab and the GLAMOR Lab under the guidence of Professors Erdem Biyik and Jesse Thomason.

I am currently exploring methods of increasing sample efficiency in generalized reinforcement learning algorithms for robotic tasks by integrating information abstraction/data manipulation techniques, enabling robots to learn complex tasks with fewer interactions and improved generalization. My approach is grounded in a strong mathematical foundation, particularly in probability theory, statistical inference, and information theory, which guide my understanding of uncertainty, generalization, and learning efficiency in reinforcement learning models.

Statement of Intent

I aspire to pursue a career in academia, where I aim to advance our understanding of robot and human perception. My first goal is to develop generalized learning methods with sample efficiency comparable to that of human learners, leveraging rigorous mathematical principles to optimize learning dynamics. Should this goal be achieved, I would then focus on either: 1. Further improving efficiency by minimizing computational and energy costs while maintaining intelligence, or 2. Enhancing agents’ ability to derive meaningful insights from sparse rewards.

Although I am still developing the technical depth needed to fully define my research trajectory, I am committed to strengthening my theoretical foundation in deep reinforcement learning, perception, and information theory. My current focus is on understanding state-of-the-art sample-efficient methods in robotics, where I apply principles from probability, statistics, and optimization to improve learning algorithms.

Other Interests

I am a competitive FPV drone racer and was part of the Korean National Drone Racing Team in 2021 and 2022. I am also part of the 3D4E (3D for All) organization where I focus on smooth surface modeling. I am an avid reader of LessWrong, where I engage with discussions on rationality, decision theory, and Bayesian reasoning. I read this mostly to refine my own decision making process and understand how us humans tend to behave. Whether or not this relates to my research is an open question, but I enjoy the journey nonetheless.