Hyunjun Kim - Staff Researcher at Samsung Advanced Institute of Technology (SAIT)

Hyunjun Kim

Staff Researcher at Samsung Advanced Institute of Technology (SAIT)

Seongnam-si, South Korea

Join Prog.AI to see contacts

Summary

👤

Senior

🎓

Top School

Hyunjun Kim is a Staff Researcher with 11 years of expertise building high-performance GPU kernels, optimizing NPU compiler passes, and enabling heterogeneous ML accelerators for mobile SoCs. At Samsung’s UDLC team he leads NPU work and researches AI/metaheuristic-based optimizations and analytical cost models, having enabled Exynos NPU, PIM, and GPU clusters in a production compiler. His background spans systems research on GPU memory and unified virtual memory at Sungkyunkwan University and practical ML engineering at Samsung Electronics, where he improved on-device localization accuracy and inference latency. An active open-source contributor, he strengthened Samsung’s TizenRT IoT RTOS and added dynamic-tensor shape inference to the On-device Neural Engine, reflecting a rare blend of compiler research, embedded systems, and ML deployment skills. Based in Seongnam-si, South Korea, he pairs PhD-level systems rigor with hands-on kernel and compiler tuning for real-world accelerators.

11 years of coding experience

9 years of employment as a software developer

Bachelor of Science - BS, Computer Engineering, Bachelor of Science - BS, Computer Engineering at Sungkyunkwan University

English, Korean

Github Skills (16)

neural-network10

c1710

compiler10

iot10

embedded10

compiler-compiler10

inference10

runtimes10

tizen10

c1110

sys10

shapes10

iot-devkit10

optimization10

tensorflow9

Programming languages (3)

C++CPython

Github contributions (5)

Samsung/TizenRT

Feb 2018 - Feb 2020

TizenRT is a lightweight RTOS-based platform to support low-end IoT devices

Role in this project:

Embedded Systems Engineer / IoT Developer

Contributions:85 commits, 139 PRs, 54 pushes in 2 years

Contributions summary:Hyunjun primarily focused on enhancing the TizenRT platform for IoT devices. Their commits show work on the `framework/st_things` directory, indicating contributions to the core functionality of the TizenRT operating system and related libraries. The user implemented logic to handle WiFi disconnections and reconnections, optimized memory usage, and added an internal function for shutting down the system. These changes contribute to the robustness and efficiency of the platform.

iot-devicesmqttrtosembeddedfreertos

Samsung/ONE

Jun 2020 - Nov 2021

On-device Neural Engine

Role in this project:

ML Engineer

Contributions:76 reviews, 19 commits, 33 PRs in 1 year 6 months

Contributions summary:Hyunjun contributed shape inference functionality for various operations within the "On-device Neural Engine," specifically focusing on improving support for dynamic tensors. The user implemented shape inference for "ZerosLike", "Tile", "OneHot", "SquaredDifference", "ReduceMean", "Split", and "ReduceMin" operations. This work involved modifying core files related to shape inference and adding corresponding test cases to ensure the correct behavior of these operations, improving the neural engine's ability to handle dynamic tensor inputs.

on-device-aion-deviceoptimizationruntimeneural-engine

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.

Request Free Trial