Shane (Sanghyun) Son

About

Ph.D. in ECE from Seoul National University (Advisor: Kyoung Mu Lee), where I developed EDSR (9,900+ citations) for deep learning-based image restoration. At Samsung Electronics (AI Center), I built and deployed domain-specific LLM systems to 10,000+ internal engineers, trained models up to 70B parameters, and improved GPU utilization ~2x. Now at TwelveLabs, working on video understanding and multimodal agents.

Experience

• Machine Learning Research Engineer at TwelveLabs
  Sep 2025 - Present · 11 mos

• 삼성전자 ()
  • AI Research Engineer
    Apr 2024 - Sep 2025 · 1 yr 6 mos

    - Prototyped an R&D copilot for semiconductor researchers (May 2025 - Sep 2025). - Developed a general-purpose reasoning model for manufacturing process automation (Dec 2024 - Apr 2025). - Led optimization of a distributed RL system scaling to 128+ GPUs to train 70B-parameter reasoning models (Dec 2024 - Apr 2025). - Developed an LLM-based summarization system for 10,000 internal engineers (Jul 2024 - Dec 2024). - Led development of an LLM-based safeguard and orchestration system on a cross-functional team; received SWE Best Practice Award (Aug 2024 - Dec 2024). - Led development of a multimodal data-processing pipeline for multi-modal LLM (Apr 2024 - Jul 2024). - Best Collaboration Award from the AI Research Center (Top 1% performer, Aug 2024). - Built an internal GPU resource-management tool as a solo project; adopted as a center-wide top 1% best practice. (Dec 2024 - May 2025).

  • Computer Vision Research Scientist
    Sep 2023 - Mar 2024 · 7 mos

    - Researched and prototyped a deep learning-based ISP pipeline to improve low-light detection for automotive machine vision (Oct 2023 - Apr 2024).

• Student Research Internship at 구글
  Jan 2019 - Jun 2019 · 6 mos

  Mentor: Ming-Hsuan Yang Research topic: Real-world image super-resolution by generating realistic image pairs.