Hello, I’m Minu.
I’m a cofounder and the Chief Science Officer at Southbridge. I work on reliable agents, difficult data, and the long stretch between a good model and useful work.
I grew up in Yongin, South Korea, and moved to Singapore at nineteen to study history and physical sciences at Yale-NUS College.
I chose Yale-NUS because I wanted to move between old texts and new tools without having to pretend they were unrelated. Some days I learned Latin and Greek. On others I ran simulations, worked with archives, or tried to make machine-learning models distinguish Roman coins struck from the same hand-carved die.
The coin project stayed with me for four years. We gathered and standardized thousands of auction-house photographs so that dies could be compared at scale. The larger question was economic: if we could estimate how many dies were used, we could begin to estimate Roman mint production and money supply. What I remember most clearly is how much work had to happen before there was anything clean enough to analyse.
Research
My academic work ended up spanning physics, history of science, and computational research. I ran numerical simulations for a 2022 Physical Review Research paper on microscopic thermalization and ion transport. At the same time, I wrote my honours thesis on Wolfgang Pauli’s correspondence with Carl Jung and the way his search for unity responded to the historical realities of the twentieth century.
Later, at Nervotec, I worked on a 200-patient study at Singapore General Hospital using remote photoplethysmography to estimate blood pressure and hemoglobin from video. The study was published in JMIR Formative Research. My contributions covered software, analysis, validation, investigation, and review.
From research to products
After college I wanted to see what happened when research met actual users, deadlines, and regulation. I started in CRM consulting at HubSpot, then joined Nervotec. Alongside the clinical study, I built a Notion-based quality-management system and prepared the documentation for the company’s EN ISO 13485 certification. I eventually managed work across engineering, design, and clinical teams.
At Meta, I worked as a Technical Solutions Consultant on the Korean-market launch of Collaborative Ads. I sat between the product team and Korean advertisers, working on signal, measurement, and delivery. It was a short chapter, but it made the problem of data crossing company boundaries very concrete.
Southbridge
I cofounded Southbridge with Hrishi Olickel in 2025. We build agents that can work through large, messy collections of files for hours without quietly losing the thread. Our open-source runtime, Hankweave, makes those workflows repeatable and repairable.
My history training is more useful here than I expected. I tend to make agents write down their plans, intermediate findings, and definitions of success before they act. It is close to the way I was taught to work with an archive: preserve provenance, separate evidence from interpretation, and make the argument defensible.
Korea
Korea has remained part of my work even while I have lived abroad. At fifteen, I helped translate a 163-page guide to Namhansanseong into English and handed it out to visitors in Insadong. At Meta, I worked on a product launch for Korean advertisers. In June 2025, I returned to Gangnam to teach a four-hour enterprise AI seminar in Korean, with support from Anthropic’s team.
Outside work, I still care about old languages, archaeology, theatre, and singing. I have excavated at the Roman port of Sanitja in Menorca, worked with rare materials at Singapore’s National Library, and performed in plays, a cappella concerts, and an original musical. I expect some of those things will find their way into the writing here.