Multimodal AI Research & Engineering
Based in Seoul, Korea
Contact: yoon7k3@gmail.com
I am a researcher and engineer focused on multimodal language models. My work spans vision, speech, and language, with an emphasis on turning ambiguous real-world interaction into measurable and useful model behavior.
MemEIC: A Step Toward Continual and Compositional Knowledge Editing
Re-engineered MEND for sequential multimodal editing across two VLM backbones; built per-edit diagnostics, isolated low-efficacy and NaN failure modes, and stabilized repeated updates for reproducible evaluation.
* Equal contribution.
Vision and Learning Lab, Seoul National University
Research Intern
Apr 2025 – Jun 2026
Built scalable data-generation, annotation, and evaluation pipelines for two multimodal benchmarks, and conducted model failure analysis to refine task and dataset design.
Electronics and Telecommunications Research Institute
Research Intern
Jan 2025 – Feb 2025
Investigated continual and compositional knowledge editing in a sequential multimodal setting and implemented the MEND baseline used in MemEIC’s cross-backbone experiments.
Asan Medical Center
Research Assistant
Jun 2023 – Aug 2023
Built a medical-document de-identification pipeline combining OCR and LLM-based entity detection to locate and mask personally identifiable information.
Sungkyunkwan University
B.S. in Data Science
2026 · GPA 4.45 / 4.5 · Ranked 1st in the College of Computing and Informatics