Junhee Yoo
Principal Engineer at Samsung Electronics
South Korea
Summary
🤩
Rockstar🎓
Top SchoolJunhee Yoo is a Principal Engineer and system IP performance architect at Samsung Electronics with over 8 years of experience shaping Exynos mobile SoC architecture. Holding a PhD from Seoul National University, he blends deep computer-architecture research with hands-on engineering to optimize processor subsystems and system-level performance. Beyond silicon, he contributes code as a part-time developer—most notably improving performance and multi-GPU support for the high-profile open-source Leela Zero Go engine by introducing fp16 weight storage and memory/layout optimizations. Known for fixing low-level correctness and compilation issues as well as building OpenCL schedulers, he bridges hardware constraints and software efficiency in production-grade systems. Based in South Korea, he combines academic rigor with practical delivery in high-stakes consumer SoC projects.
8 years of coding experience
Doctor of Philosophy - PhD, Electrical, Electronics and Communications Engineering, Doctor of Philosophy - PhD, Electrical, Electronics and Communications Engineering at Seoul National University
Github Skills (11)
opencl10
gpu-programming10
c-language10
cprogramming-language10
performance-optimization10
tensorrt9
tensor9
machine-learning8
game-dev8
game-development8
game-engine8
Programming languages (4)
Github contributions (5)
leela-zero/leela-zero
Nov 2017 - Oct 2019
Go engine with no human-provided knowledge, modeled after the AlphaGo Zero paper.
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:41 commits, 60 PRs, 6 pushes in 1 year 10 months
Contributions summary:Junhee's contributions primarily focused on optimizing the Leela Zero Go engine. They implemented the use of half-precision (fp16) storage for neural network weights to improve performance on certain GPUs by reducing memory bandwidth usage. They also addressed incorrect size calculations and compilation errors related to the half-precision implementation, as well as introduced multi-GPU support for playing games, including the creation of an OpenCL scheduler and worker threads. Further optimizations were made to UCTNode memory layout and the performance of core calculations within the engine.
golangalphago-zerozeroalphagoknowledge
ihavnoid/leela-zero
Jul 2018 - Oct 2019
Go engine with no human-provided knowledge, modeled after the AlphaGo Zero paper.
Contributions:100 commits, 2 PRs, 147 pushes in 1 year 2 months
golangalphago-zerozeroalphagoknowledge
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.