Sungjun Kim
Machine Learning Platform Engineer
Seoul, South Korea
Summary
🤩
Rockstar🎓
Top SchoolSungjun Kim is a Machine Learning Platform Engineer with 10 years of experience building production ML systems and leading platform teams for companies like LINE and Coupang. He combines end-to-end expertise—from data engineering and feature stores to real-time recommenders, multi-armed bandits, and observability—to make data science reliably consumable as production services. Sungjun has steered company-wide ML platform design, integrating Airflow, MLflow, BentoML and Ray, and led Timeline recommendation and global operations efforts at scale. An active open-source maintainer on projects such as Clipper and BentoML, he has deep hands-on knowledge of prediction-serving stability, dependency management, and metric-driven operations. Based in Seoul, he brings a rare mix of embedded-to-cloud experience and a pragmatic focus on operationalizing models rather than just building them.
10 years of coding experience
16 years of employment as a software developer
bachelor's degree Computer Science, bachelor's degree Computer Science at Yonsei University
Stackoverflow
463reputation
54kreached
11answers
0questions
Github Skills (21)
c-language10
back-end-development10
cmake10
cprogramming-language10
boost10
devops10
dockers9
python9
docker9
prometheus8
conda8
system-design8
stackview6
azure6
macos6
Programming languages (10)
Github contributions (5)
ucbrise/clipper
Nov 2017 - May 2018
A low-latency prediction-serving system
Role in this project:
Back-end & DevOps Engineer
Back-end & DevOps EngineerContributions:11 commits, 64 PRs, 8 pushes in 6 months
Contributions summary:Sungjun primarily focused on enhancing the Clipper prediction-serving system, addressing build dependencies and system stability. Their contributions included updating CMake configuration to support Boost 1.66.0 and 1.67.0, fixing a bug in the query frontend, and resolving issues related to conda environment management and dependencies. Furthermore, they implemented features for metric monitoring, including adding and deleting model containers from the Prometheus configuration. They also fixed an infinite loop of GarbageCollectionThread.
pythonservingpredictiondeep-learninglatency
withsmilo/clipper
Oct 2017 - Jun 2019
A low-latency prediction-serving system
Contributions:206 pushes, 109 branches in 1 year 8 months
pythonservingpredictiondeep-learninglatency
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.