Shane Son
Machine Learning Research Engineer at TwelveLabs
Seoul, South Korea
Summary
🤩
Rockstar🎓
Top SchoolShane Son is a Machine Learning Research Engineer based in Seoul with nine years of experience bridging academic research and industrial-scale AI systems. He holds a Ph.D. from Seoul National University where he specialized in image super-resolution and developed practical resampling models, and he contributed open-source PyTorch implementations (notably an EDSR port) that reflect strong emphasis on reproducibility and optimization. At Samsung he led multimodal data pipelines, LLM-based summarization and safeguards, and scaled distributed RL training to 128+ GPUs for 70B-parameter reasoning models, earning top collaboration and best-practice recognitions. Now at TwelveLabs, he focuses on applied ML research and production-ready model engineering, combining deep CV expertise with large-model orchestration. Colleagues would describe him as a pragmatic researcher who turns state-of-the-art ideas into documented, usable tools for cross-functional teams.
9 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy - PhD, 전기정보공학, Doctor of Philosophy - PhD, 전기정보공학 at 서울대학교 (Seoul National University)
Github Skills (10)
neural-network10
pytorch10
convolutional-neural-networks10
deep-learning10
trainings10
python10
image-processing10
modeling10
super-resolution10
machine-learning9
Programming languages (1)
Github contributions (5)
sanghyun-son/EDSR-PyTorch
Sep 2017 - Aug 2019
PyTorch version of the paper 'Enhanced Deep Residual Networks for Single Image Super-Resolution' (CVPRW 2017)
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:95 commits, 12 PRs, 77 pushes in 1 year 11 months
Contributions summary:Shane's contributions focus on the development of the EDSR PyTorch model. They initiated the project, created utility functions, and established the basic structure for training and evaluation. The user made changes to the training scripts, implemented memory-efficient forward passes, and incorporated techniques like self-ensembling, demonstrating a solid understanding of model optimization and performance. They also updated the README and included example demo scripts, suggesting efforts towards usability and documentation.
pytorchimage-manipulationresidual-networksdeep-learningsingle-image
limbee/NTIRE2017
Feb 2017 - Sep 2017
Contributions:247 commits, 26 pushes, 2 branches in 6 months
pytorchresidual-networksdeep-learningsingle-imageresolution
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.