Sang-ik Cho
Director, Research Product Architecture at Meta
Redmond, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolSang-ik Cho is a seasoned research product architect and engineering leader with nine years of industry experience and a PhD in Acoustics, currently directing Research Product Architecture at Meta in Redmond. He blends deep systems and acoustics expertise with hands-on backend and DevOps work—contributing to high-profile open source projects like Ray and vLLM where he improved control planes, CI robustness, and inference scheduler performance. Known for turning research into production-grade systems, he has driven cross-disciplinary teams across audio, systems integration, and large-scale ML infrastructure. A former senior committer on Ray-related projects and an MTS at xAI, he brings both rigorous academic training and pragmatic engineering sensitivity to reliability, performance, and maintainability.
9 years of coding experience
18 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Acoustics, Doctor of Philosophy (Ph.D.) Acoustics at Penn State UniversityBachelor of Engineering (B.E.) Electrical Engineering, Bachelor of Engineering (B.E.) Electrical Engineering at The Cooper Union for the Advancement of Science and Art
English, Korean
Github Skills (18)
docker10
c-language10
python10
multithreading10
bash10
dockers10
cicd10
scheduler10
llm10
error-handling10
ray10
cprogramming-language10
grpc10
testing9
inference9
Programming languages (8)
Github contributions (5)
ray-project/ray
Jan 2020 - Jan 2023
Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:3 releases, 6136 reviews, 781 commits in 3 years
Contributions summary:Sang-ik primarily contributed to refactoring the code to improve the performance of various internal APIs such as `TryReadObjectRefStream`. They made changes to the handling of stream operations within the GCS, particularly when dealing with edge cases such as out-of-order writes, implementing mechanisms to maintain data integrity. Furthermore, the user modified the exception handling for the same method, which made better error messages.
pythonconsistsruntimetensorflowserving
vllm-project/vllm
Dec 2023 - Dec 2024
A high-throughput and memory-efficient inference and serving engine for LLMs
Role in this project:Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:602 reviews, 136 PRs, 25 pushes in 11 months
Contributions summary:Sang-ik integrated the Ray framework for optimizing the control plane of the LLM engine, improving its efficiency. They added tests for the block manager and scheduler, enhancing the system's reliability. The user also refactored and optimized various parts of the scheduler and the model runner, including chunked prefill data updates, and attention metadata handling, demonstrating a focus on performance and efficiency. Furthermore, they made significant modifications to the CI and documentation aspects, showing DevOps knowledge.
amdcudadeepseekgpthpu
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.