Sooheon Kim is a Head of Strategy with 11 years of machine learning and AI experience, currently focused on training virtual cells using in-house proteomics at PanThera. He blends hands-on ML engineering — from Korean NLP and productionized inference pipelines to graph neural network contributions in JAX — with strategic leadership across academia, startups, and public sector projects. His background spans drug discovery AI, cancer research, and applied ML for government and ad-tech, giving him a rare perspective on both scientific and product-facing ML workflows. An active contributor to the google-deepmind jraph library, he implemented attention normalization parameterization and rigorous tests, signaling a commitment to robust, reusable ML tooling. Fluent in cross-disciplinary collaboration, he also brings international experience from service as an interpreter with NATO-awarded deployment.
11 years of coding experience
5 years of employment as a software developer
Bachelor's degree, International Studies / Business, Bachelor's degree, International Studies / Business at Korea University
Contributions:3 reviews, 5 commits, 2 PRs in 1 day
Contributions summary:Sooheon contributed to the JAX-based graph neural network library by adding new utility functions for segment-wise variance and normalization operations. They also introduced a parameterization for the attention normalization function, enabling the use of alternatives to the default softmax. Furthermore, the user implemented tests to ensure the correctness of these new functionalities and provided code formatting fixes.
A Clojure library for deconstructing Korean unicode syllable characters into alphabet characters
Contributions:5 commits, 1 PR, 29 pushes in 2 years 2 months
alphabetclojure-librarykoreanclojureunicode
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.