Hi 👋, I'm Stuart. I aim to be both an entrepreneur and a dedicated engineer. I develop new things, optimize systems, and build teams. I truly love the feeling of tackling new challenges from scratch. I co-founded Blux (formerly Z.Ai), an AI startup specializing in empowering Korean e-commerce through AI-driven personalization. Blux raised $3M+ and is personalizing over 10 million Korean users' online journeys monthly. I am now pursuing my Master of Science degree in Computer Science at Stanford University. I'm currently focusing on MLSys research @ Hazy Research (Prof. Chris Ré). Starting in June 2025, I'll be working as a machine learning engineer @ Cursor, building new AI kernels to make Cursor's 4,000+ GPU cluster go brrr. In addition, I professionally play electric guitar and compose music. I have released two albums and have been the lead guitarist and producer for various rock bands since 2008.
Education
Stanford University
M.S. in Computer Science
- Recipient of the merit-based Kwanjeong scholarship ($60,000)
Seoul National University
B.S. in Computer Science and Engineering
- GPA: 4.21/4.3 (class rank: 1st)
- Recipient of the National Science and Engineering Scholarship for Gifted Students (full scholarship)
- Member of the College of Engineering Honor Society
Experience
Stanford AI Lab (SAIL)
Research Assistant @ Hazy Research (Prof. Chris Ré)
- Designing speedy multi-GPU kernels for AI
- Working on ThunderKittens (recent blog post)
Blux (formerly Z.Ai)
Co-Founder and CTO
- Real-time recommender systems for e-commerce
- Now serves 10M+ monthly end-users with $1M+ in annual revenue
- Built and led a team of 15+ engineers.
- Built lots of recommender models, multi-tanent MLOps architecture (Kubernetes was fun), microservices for data collection and model serving (thank you FastAPI), SDK in 7 programming languages, etc.
Architecture and Code Optimization Lab, Seoul National University
Research Assistant (Prof. Jae W. Lee)
- Designed and implemented a novel embedding clustering algorithm in C++, reducing the main memory access by up to 44% in commercial deep learning recommendation models (DLRMs).
- Resulted in a paper accepted at ASPLOS 2021.
Music and Audio Research Group, Seoul National University
Research Assistant (Prof. Kyogu Lee)
- Implemented an audio processing architecture using CNN, cGAN, super-resolution, and the Griffin-Lim algorithm, contributing to a commercial singing voice synthesis project.
- Performed data labeling on audio and MIDI data with millisecond-precision using Logic Pro and Python, preparing 30+ songs for model training (was fun, but very painful)
US Army
Sergeant
- Served as a Combat Medic (68W) in a U.S. Army cavalry unit for 21 months.
- This was possible through the Korean Augmentation to the United States Army (KATUSA) program which fulfilled my mandatory military service requirement as a South Korean.
Projects
ThunderKittens
Open Source Research Project
- Helps you write speedy GPU kernels for AI
- Extended the original implementation to support multi-GPU kernels
- Currently working on a new GPU-side virtual machine project!
ELF32 Dynamic Linker for Raspberry Pi
Personal Project
- Ever wanted to run a dynamic linker on bare metal, and port shared libraries?
- On May 22, 2025, I taught this as a lecture/lab at Stanford University (CS 240LX: ELF and Dynamic Linker)
SampyoNet
University Team Project
- Developed a TensorFlow-based deep learning model and React.js-based mobile application for automated gravel quality assessment in concrete manufacturing.
MERCI
Open Source Research Project
- Fast embedding reduction algorithm for deep learning recommendation models (DLRMs) and other systems with extremely large-scale embedding tables.
LLVM Compiler Optimization
University Team Project
- Achieved 2nd place among 13 teams in an LLVM optimization competition at Seoul National University.
Homemade Neural Network
Personal Project
- A deep learning framework written using only native Python (not even NumPy).
Five-in-a-row with AI
Personal Project
- A game of five-in-a-row, with GUI and an AI opponent.
- GUI and AI written in native Python.
- AI implemented using minimax algorithm.