Systems researcher, Department of Computer Science and Engineering, Seoul National University
I study how numerical algorithms can be redesigned around modern GPU hardware, especially Tensor Cores, memory hierarchy, and data movement constraints. My work focuses on high-performance GPU computing, numerical algorithms, and hardware-aware optimization.
My recent work focuses on GPU-accelerated numerical computing, including Tensor Core FFT kernels, GPU-aware sparse direct solvers, and multi-GPU solver components. Across these projects, I focus on reducing data movement, improving locality, and achieving predictable performance on modern accelerators.
Accelerating Fast Fourier Transform with Tensor Cores.
, S. Bae, S. Lee, J. Lee, J. Kim, D. Kim, K. Ahn, and J. Lee.
PPoPP 2027 (accepted).
SEDD: Scalable and Efficient Dataset Deduplication with GPUs.
Y. Son, , and J. Lee.
KDD 2026. Paper
SnuSOLVER: Optimizing Sparse Direct Solvers for Heterogeneous Systems.
, J. Lee, J. Kim, D. Kim, K. Ahn, H. U. Cho, S. Baek, and J. Lee.
ICS 2025.
Efficient Homomorphic Max Algorithm for Multivariables.
J. H. Cheon, W. Cho, S. Hong, and .
Preprint, 2020.
Accelerating 2D Convolution using cuFFTDx.
S. Bae, J. Lee, , S. Lee, and J. Lee.
Korea Computer Congress (KCC), 2025.
New Adaptive Matching Order and Performance Comparison for Subgraph Matching Problem.
S. Min, W. Shin, , and K. Park.
KCC Best Paper Award, 2021.
GPU-accelerated numerical computing
My recent research redesigns numerical algorithms for modern GPU hardware. I have worked on Tensor Core FFT kernels, GPU-aware sparse LU factorization, and multi-GPU sparse solver components, with an emphasis on reducing data movement, improving locality, and matching computation to GPU memory hierarchy and Tensor Cores.
I also explored memory-aware quantum circuit partitioning as a lower priority direction, focusing on locality and memory traffic reduction.
Large-scale LLM data systems
I contributed to large-scale crawling and preprocessing pipelines, developed a GPU-accelerated LSH-based deduplication framework, and profiled training efficiency with CUPTI and Nsight.
Earlier research
As an undergraduate researcher in Thunder Research Lab, I developed GPU kernels and simulations in OpenCL, CUDA, and PyTorch. I also worked on DAF graph matching with Prof. Kunsoo Park and CKKS homomorphic encryption optimization with Prof. Jung Hee Cheon.
Seoul National University, Department of Computer Science and Engineering
M.S. in Computer Science and Engineering, Sep. 2022 - Aug. 2025
Advisor: Prof. Jaejin Lee
Seoul National University, Department of Computer Science and Engineering
B.S. in Computer Science and Engineering, Mar. 2016 - Aug. 2022
Minor in Mathematics
Seoul Science High School
Mar. 2013 - Feb. 2016
Teaching Assistant, Accelerator Programming School
Teaching Assistant, High Performance Parallel Computing
Teaching Assistant, Algorithms
Teaching Assistant, Korean Mathematical Olympiad Winter Camp