dev.log

A place for peace of mind

Sangwu Lee

Hi, I'm Sangwu. I work on production LLM/RAG systems at Quizard @ NYC. While I'm not working, I work on multimodal learning research and train ML models on anime datasets. I believe that ML research "results" and "process" should be transparent and reproducible. I try to use this place to keep track of my own research progress.

I grew up in South Korea, Netherlands, and New Mexico. My first programming language was in C, but began seriously coding with JS/TS for fullstack development since 2015. After being inspired by Deepmind's progess in Machine Learning, I began studying ML in 2018 starting with NLP and Recommender systems. As for my research philosophy, I believe persistence and iteration speed are the key. I try to keep up with latest ML systems and hardware research to improve my training systems.

During my undergraduate years, I've worked on various research projects in machine learning / Human computer interaction with a focus on multimodal learning. Together with ROC-HCI Lab, I've worked on developing Health AI systems for diagnosing Parkinson's disease and movement disorders using multimodal data. I've still an active collaborator of CMU's Language Technologies Institute where I work on multimodal reasoning tasks with vision language models. I mostly publish in ACL, AAAI, ACII, Nature Digital Medicine, and my favorite ArXiv

Apart from ML research, I still consider myself a mathematian at heart. Specifically, my interest is in analysis with a focus on ODE/PDE and probability theory. I'm currently plowing through Axler's Measure theory book and Le Gall's probability book.