Founding AttentionX

Harvard’s graduate application has an “Additional Materials” section. I figured I could talk about my experience orthogonal to academia and research but still very meaningful to me and where I spent a lot of time and life force on over the past 2 years.

I would like to briefly share another part of my research journey that is not fully captured by my CV or statement of purpose: AttentionX, an AI research community I founded in Seoul.

I returned to Korea in 2024 for two years of mandatory alternative military service. Although I was born in Korea, I had lived in the United States since I was two years old, so I returned without an existing academic or research network. Rather than treat those two years as a pause from the research trajectory I had begun at Berkeley, I decided to build the environment I wanted myself.

I started AttentionX by recruiting an initial cohort of 20 undergraduate and graduate students in computer science and mathematics from Seoul National University and KAIST through student forums. I reached out to companies for compute sponsorship, recruited faculty collaborators, invited researchers from academia and industry, and organized the group into four research agendas: LLM reasoning, multimodal learning, 3D vision, and diffusion models.

Over the next two years, AttentionX grew into a research community that produced eight papers accepted at venues including NeurIPS, ECCV, ICML, CVPR, and EMNLP in collaboration with four SNU professors and two KAIST professors; I coauthored five of these projects. We also built a broader intellectual community around the group. AttentionX co-hosted AI hackathons and events with OpenAI, DeepMind, and Anthropic, which also provided API credits, and hosted researchers from both industry and academia, including former Harvard professor Julia C. Lee for a discussion on AI for climate and scientific discovery.

What surprised me most was how much the environment around research mattered. Research progress was not only a function of technical ability. Who sat next to whom, how projects were grouped, how freely people exchanged unfinished ideas, what resources were available, how the office itself was organized, and whether researchers felt ownership over the problems they pursued all changed what people attempted and ultimately produced.

That realization has shaped how I think about leadership. I had the unusual opportunity to participate in nearly every layer of building AttentionX: recruiting people, securing resources, designing the physical and intellectual environment, organizing events, connecting researchers, and pushing projects toward completion. Four teams that emerged from the community later went on to Y Combinator, and other members built companies of their own.

AttentionX changed my understanding of what it means to contribute to research. I do not only want to produce strong work myself. I want to build environments where ambitious researchers and founders find one another, develop their own research taste, and pursue ideas that might not have existed otherwise. That instinct to create the research environment rather than wait for one to appear is something I hope to continue throughout graduate school and eventually in the research organization and company I hope to build.