Jeonghyun Kim
M.Sc
AI Engineer @ SNOW Corp.
Contact: mildjay16@gmail.com
Jeonghyun Kim, M.Sc, is an AI engineer at SNOW. He obtained his bachelor's degree at Chung-Ang Univ. (CAU) in Advertising & Public Relations and a double major in Technology Art. After that, he obtained his master's degree in 3D computer vision/graphics from Graduate School of Culture Technology, KAIST, under supervision of Prof. Woontack Woo (GSCT) and Prof. Minhyuk Sung (SoC). In graduate school, he tackled 3d computer vision/graphics problems related to 3D modeling and reconstruction by investigating compositional understanding on 3D data for XR through AI.
His current focus lies in solving real-world AI challenges, leveraging his extensive hands-on experience in developing and productionizing end-user AI services. He is particularly skilled in large-scale GPU cluster operations for production-grade AI applications. As an AI engineer, he strives to deliver AI to the world with enhanced efficiency and reliability.
Experiences
SNOW Corp.
2023.12 - (Present)
AI Engineer
Building an On-premise Kubernetes GPU Cluster from scratch including monitoring system.
Enhancing GPU utilization on Kubernetes, through workflow orchestration and GPU sharing strategies.
Managing a Large-scale Kubernetes GPU Cluster with hundreds of GPUs to power generative AI services for SNOW, EPIK, and other applications.
Gave an oral presentation titled “Migrating to Kubernetes GPU Cluster for Large-scale AI Service Operations” at NAVER ENGINEERING DAY 2024, an internal technical conference.
presentation video (spoken in Korean)
Super Labs, NAVER
2022.12 - 2023.12
AI Engineer
Research on image/video/3D generation tasks based on 3DMM, NERF, GAN,diffusion model, etc.
Developing generative model APIs using open-source libraries and implementing off-the-shelf research papers
Building CI/CD infra structures to serve large scale AI models on cloud computing platforms using k8s, helm, and terraform
Serving AI models with GPU optimization on code/infra level
Gave a talk on “Replicating Interactive 3D Scene using Shape & Structural Priors” at Korea Institute of Science and Technology (KIST)
etc: diffusers contributor
Ubiquitous Virtual Reality (UVR) Lab., KAIST
2020.03 - 2022.08
Student Researcher
Projects
Development of Open XR platform for high immersive collaboration (2022.03 - 2022.08)
- National Research Council of Science and Technology(NST) funded by the Ministry of Science and ICT(MSIT)
Dynamic Digital Twin for Realistic Untact XR Collaboration (2021.03 - 2022.08)
- National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT)
Publications
Seg&Struct: The Interplay between Part Segmentation and Structure Inference for 3D Shape Parsing
Jeonghyun Kim, Kaichun Mo, Minhyuk Sung*, and Woontack Woo*
(*Co-corresponding Authors)
IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2023 (Algorithm Track)
Education
Master of Science, Graduate School of Culture Technology, KAIST
2020.03 - 2022.08
3D Computer Vision/Graphics, Deep Learning, AR/VR
Supervisor: W. Woo (GSCT) and M. Sung (SoC)
Thesis: The Interplay between Part Segmentation and Structure Inference for 3D Shape Parsing
Scholarship: CT-AR Program
Bacheler of Arts, Chung-Ang University
2014.03 - 2020.02
Major in Advertising & Public Relations
+ Double major in Technology Arts
Other Experiences
Seoul Metropolitan Police Agency (SMPA)
2015.02 - 2016.11
Served Military Service as Auxiliary Police