profile
Seunghyuk Oh
CV
profile
Seunghyuk Oh
Master's Student in Artificial Intelligence

I am a Master's student at KAIST ALIN Lab, specializing in AI under Prof. Jinwoo Shin.

Previously, as a co-founder of a startup, I honed my skills as a software engineer, using technology to tackle complex challenges. This entrepreneurial journey ignited my passion for artificial intelligence, leading me to transition into AI research. Now, I’m dedicated to contributing to the advancement of AI by merging my practical experience with academic research.

Projects

  1. ReVISE Research
    Fine-tune LLMs to self-verify and self-refine.
    2024.12 ~ 2025.02
    • Developed a framework that enables LLMs to refine reasoning at test time through intrinsic self-verification.
    • Introduced curriculum learning to improve the verification performance of models that play a key role in the overall framework.
  2. OCTree Research
    Optimizing column feature generator with decision Tree reasoning.
    2024.01 ~ 2024.05
    • Developed a LLM-based optimization framework on tabular domain utilizing decision tree reasoning.
    • Working to make OCTree an open source library.
    https://arxiv.org/abs/2406.08527
  3. HOMER Research
    Hierarchical context merging for long context modeling.
    2023.08 ~ 2023.09
    • Developed a training-free, efficient long-context modeling technique leveraging a divide-and-conquer strategy.
    • Proposed an attention score-based token selection algorithm to find important tokens that will not be pruned during context compression.
    https://arxiv.org/abs/2404.10308
  4. iN!T Development
    A senior developer evaluation-based recruitment solution.
    2021.12 ~ 2022.12
    • Utilized no-code tools to develop a prototype for initial hypothesis validation rapidly and focused on building internal tools to support service operations.
    • Developed front-end applications using NextJS, catering to juniors, seniors, and companies.
    • Built a backend server with NestJS to automate tasks previously handled manually by the operation team.
    https://www.venturesquare.net/859963
  5. MEDIAR Research Development
    A computer vision-based lung cancer diagnosis solution.
    2021.02 ~ 2021.06
    • Designed a computer vision solution to assist in the lung cancer diagnosis process, transitioning from traditional methods reliant on doctors' direct microscopic observation to an AI-supported approach.
    • Developed an AI system capable of analyzing high-capacity, high-resolution Whole Slide Images to identify lung cancer regions, classify the type of lung cancer in specimens, and estimate probabilities for accurate diagnosis.
    https://platum.kr/archives/160941

Work Experience

  1. Chief Technology Officer
    Weebut
    2021.01 ~ 2022.12
    • As the tech team leader, I was responsible for overseeing all technical aspects of our products.
    • Developed various business solutions, such as MEDIAR and iN!T, by implementing them as software through JavaScript-based frontend and backend development.
  2. Undergrad. Research Intern
    KAIST ALIN Lab
    2020.06 ~ 2020.12
    • Studied MirrorGAN, a model that learns the relationship between images and captions by converting a caption into an image and then back into a caption.
    • Applied semi-supervised learning to MirrorGAN.
    • Showed that semi-supervised learning can achieve comparable performance to fully supervised models.
    https://alinlab.kaist.ac.kr
  3. Undergrad. Research Intern
    KAIST AIPR Lab
    2019.12 ~ 2020.01
    • Contributed to the OpenXAIProject/stable-baseline-tf2
    • Upgraded the Deep Q-Network codebase from Tensorflow 1.0 to 2.0.
    https://ailab.kaist.ac.kr/
  4. Undergrad. Research Intern
    KAIST Data Mining Lab
    2019.07 ~ 2019.08
    • Studied matrix factorization, a classical recommendation system technique.
    • Focused on dividing the list of items into categories and making recommendations by category.
    • Measured how low performance and computational costs are compared to the recommendations made across the entire dataset.
    https://sites.google.com/view/kaistdata/home

Education

  1. M.S. in Artificial Intelligence
    Korea Advanced Institute of Science & Technology
    2023.9 ~  On Going
    Advisor: Jinwoo Shin
  2. B.S. in Electrical Engineering
    Korea Advanced Institute of Science & Technology
    2017.02 ~ 2023.08
    Minor in Computer Science and Industrial & Systems Engineering