Sungmin Kang

Graduate Research Intern @Inha University 인하대학교

Seoul, KR

Large-Scale Machine Learning Systems Lab - https://sites.google.com/view/lmls-lab- Designed a communication-efficient FL algorithm that selectively updates high-variability layers by monitoring gradient-to-weight ratio, reducing communication costs by up to 83% while maintaining model accuracy- Developed a model-agnostic fine-tuning method for LLM by employing a parameter-wise mask based on gradient-to-weight ratio, achieving performance comparable to full parameter training by tuning 0.1% of total model parameters

Sogang University

Bachelor of Science - BS, Electronic Engineering

University of Southern California

Master of Science - MS, Electrical and Computer Engineering (Machine Learning and Data Science)