Junwoo Park

M.S. student in Computer Science at UC Irvine

junwoop5@uci.edu

Portrait of Junwoo Park

My research interests lie in multimodal AI and the safety and alignment of embodied agents. I conduct research under the guidance of Dr. Liwei Jiang.

I received my B.S. in Computer Science and Engineering from Sungkyunkwan University (SKKU), where I began my research with Prof. Sujee Lee at the AI-Healthcare Lab.

Outside research, I build apps for campus communities and previously led Google Developer Groups on Campus at SKKU.

News

  • Sep. 2026 I started my M.S. in Computer Science at UC Irvine.
  • Aug. 2026 Our preprint MR-MoL is now available on arXiv.
  • Feb. 2026 Received my B.S. in Computer Science and Engineering from SKKU.
  • Dec. 2025 Our paper “MoltiTox: a multimodal fusion model for molecular toxicity prediction” has been published in Frontiers in Toxicology!
  • Nov. 2025 Our team won First Place at 2025 Uni-DTHON, an AI hackathon on vision-language tasks! (Code)
  • Oct. 2025 Featured in Sungkyunkwan University’s online magazine (SKKU Webzine).
  • Sep. 2025 Honored to be named to Sungkyunkwan University’s 2025 President’s List!
  • Aug. 2025 VisualVroom won the Grand Prize at ICT Award Korea 2025! (article)
  • Feb. 2025 Started as an undergraduate researcher at the SKKU AI-Healthcare Lab.

Publications

Selected Projects

Project SKKU Character

Project SKKU Character

Oct. 2025 – Feb. 2026

AI platform for SKKU’s mascot characters

I fine-tuned FLUX on mascot images and built a FastAPI backend for students to generate and share artwork from text prompts, supported by the university’s entrepreneurship program.

NotiSKKU app

NotiSKKU

Aug. 2024 – Nov. 2025

Student announcement app

I led a four-person team to bring campus announcements into one app, informed by a survey of 60+ students. Built with Flutter, Playwright, and Firebase; released on Google Play.

VisualVroom system

VisualVroom

Apr. 2025 – Aug. 2025

Haptic alerts for deaf and hard-of-hearing drivers

We used a Vision Transformer on audio spectrograms to jointly classify sound type and direction with 83.9% accuracy, delivering real-time haptic alerts through a smartwatch.