Jaegeun Han
Seoul, Korea, Republic of
Summary
🤩
RockstarJaegeun Han is an AI Solutions Architect with over a decade of hands-on experience designing and optimizing large-scale GPU-accelerated AI systems for data centers and production LLM pipelines. Currently at AMD, he architects and validates ROCm-based deployments and leads cross-functional programs to boost AI workload performance across GPU, CPU, and DPU platforms. Previously at NAVER he built and stabilized HyperCLOVA X production pipelines, scaled 2,000+ GPU clusters, and delivered long-context and LoRA inference enhancements for FasterTransformer. His background spans deep CUDA optimization, TensorRT tuning, and on-prem GPU cluster deployments from roles at NVIDIA, Samsung, and research labs, with open-source contributions to CUDA educational repos showing practical SGEMM and convolution optimizations. Based in Seoul, he combines systems-level engineering with customer-facing enablement to turn cutting-edge AI research into scalable, production-ready solutions.
10 years of coding experience
Github Skills (9)
cuda10
matrix-multiplication10
convolution10
gpu-programming10
parallel-computing10
cprogramming-language9
cudnn9
c-language9
cublas9
Programming languages (5)
Github contributions (5)
PacktPublishing/Learn-CUDA-Programming
Feb 2019 - Aug 2020
Learn CUDA Programming, published by Packt
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 44 commits, 5 PRs in 1 year 7 months
Contributions summary:Jaegeun contributed to a CUDA programming repository focused on learning and implementing CUDA applications. The commits demonstrate the user's focus on matrix multiplication (SGEMM) optimization and convolution operations, showcasing their understanding of parallel programming patterns. Further commits refactor and refine various CUDA implementations, including scan operations, and incorporate results validation to ensure correctness. The user also demonstrates the use of CUDA libraries and the integration of profiling tools to measure performance.
cuda
haanjack/mnist-cudnn
Sep 2019 - Aug 2020
CUDA for MNIST training/inference
Contributions:2 reviews, 9 commits, 2 PRs in 11 months
cudainferencemnistcudnn
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.