Hi, I teach machines to draw pretty pictures.

I've been playing around with computers for a while. I've played with a lot of different things ranging from frontend, infra, backend engineering, ..., but my current focus is ML. My research is on image/video generation and creative interfaces.

I spent last few years training various image generation models ranging from GAN, VQVAE, AR, and Diffusion models. Nowadays, I've especially interested in diffusion model postraining, applying SAEs on image models, AR models, and image tokenizers.

Prior to devoting my time to image generation, I worked on multimodal learning research with CMU LTI. I've also done a fair amount of medical AI research with ROC HCI lab.

I grew up in South Korea, Netherlands, and New Mexico. I learned C when I was 11, but was quickly disappointed that it wasn't a good language for making visual mediums. Fortunately, I found my way around web development and had a lot of fun building static sites using HTML, CSS, and JQuery.

During early years of high school, I experimented with React, Angular, Laravel, Wordpress, Ruby on Rails, and Meteor.js. But, when I learned about Netflix Prize, Kaggle, and AlphaGo, I quickly shifted my focus on ML.

I recently finished my undergraduate degree in CS and Mathematics from University of Rochester. Apart from academic ML research, I spent a lot of time studying mathematics especially in real analysis. I wish I had more time to study fourier analysis and probability theory. I'm trying to teach it to myself when I can find time.

Currently, I'm working at krea.ai where I teach machines to draw pretty pictures. Previsouly, I was founding engineer at Oleve where I helped scale two consumer products from zero to a few million users in less than a year.