Seoyoon Yum

CS @ KAIST

I am a Computer Science undergraduate at KAIST. I work on efficient AI systems, focusing on LLM inference, resource scheduling, and quantization. I am currently working with Prof. Sehoon Kim as a Research Intern in SEAL Lab at KAIST. Previously, I was a research intern at UC San Diego, advised by Prof. Falko Kuester.

Seoyoon Yum

Publications

View publications

Technical Report

  1. 2026

    Zero-Shot Time-Series Foundation Models for Infrastructure Anomaly Detection

    Seoyoon Yum

    Technical Report

    Paper

View technical reports

Research Experience

  1. Sep 2026–Present

    KAIST · SEAL Lab

    Research Intern · Prof. Sehoon Kim

    NVFP4 W4A4 Quantization for Looped Transformers — Researching quantization performance in looped transformer models and optimizing inference kernels to improve speedups.

  2. Jun–Aug 2026

    University of California, San Diego

    Research Intern · Prof. Falko Kuester

    Infrastructure Anomaly Detection — Evaluated zero-shot anomaly detection with a time-series foundation model on real-world operational data, identifying limitations of forecast residuals for genuine failure prediction.

  3. Jan 2026

    KAIST · SGVR Lab

    Undergraduate Researcher · Prof. Sung-Eui Yoon

    Vision-Language-Action Models — Evaluated the semantic robustness of vision-language-action (VLA) policies in simulation under meaning-preserving instruction changes; average success fell from 60% to 25%.

  4. August 2023

    CJ AI Center

    Research Intern

    EEG-to-Robotic-Arm Interface — Built a real-time brain–computer interface mapping EEG signals to robotic-arm commands using PyTorch.

View full research experience