Structure-grounded multimodal reasoning: Developed retrieval and reasoning methods that preserve document structure — layout, tables, figures, and cross-page dependencies — turning noisy multi-page documents and video into grounded evidence for QA.
First-author work: Led first-author papers at NeurIPS, EMNLP, ACL, and CVPR on agent memory-write admission, page-grounded evidence graphs, hierarchical retrieval, multimodal document dependency parsing, and structure-preserving chunking.
Hybrid reasoning systems: Built QA pipelines combining LLMs/LVLMs, knowledge graphs, symbolic rules, and evidence filtering for accurate, interpretable answers.
Applied validation: Validated multimodal RAG in insurance, manufacturing, and healthcare, including a Samsung Fire & Marine MVP that cut manual document search time by about 40%.
Korean dependency parsing: Developed neural-symbolic dependency parsing methods that integrate morphological and syntactic information for robust Korean language understanding.
Table and document QA: Designed pipelines that convert semi-structured tables and documents into structured representations for question answering and interpretable reasoning.
Applied collaborations: Contributed to Korean TTS prosody prediction with KT AI2XL and additional projects in finance QA, sentiment-based marketing, and recommendation systems.
Thesis:A Neural-Symbolic Model for Overcoming Deep Learning Limitations in Korean Dependency Parsing
Selected outcomes: Developed neural-symbolic Korean dependency parsing and table/document QA methods; received Outstanding Paper and Student-led Research Project awards.
B.S. in Information and Computer Engineering - Pusan National University
Mar 2016 - Feb 2021 | Busan, Republic of Korea
Undergraduate Research: Prof. Jeongu Kim’s group and Prof. Hyuk-Chul Kwon’s AI Research Lab
Selected outcomes: Worked on knowledge-graph-based Korean NLP, Korean grammar correction, Korean TTS, and transformer-based Korean language models; received a merit-based Milestone Scholarship.
Honors & Awards
Deputy Prime Minister and Minister of Education Award, Competition for Best Practices in Academia–Industry CooperationNRF · 2024
Outstanding Paper Award, Korea Institute of Information and Communication Engineering (KIICE)KIICE · 2023
Grand Prize, Samsung Fire & Marine Insurance LLM-based MVP Model Idea CompetitionSamsung · 2023
Excellence Award (Technology Track), KU-Grant ProgramKorea University · 2023
Outstanding Paper Award, Korea Software Congress (KSC)KSC · 2022
Student Research Award, Student-led Research Project (HCLT 2022 paper), Graduate School of AIPusan National University · 2022
Milestone Scholarship (Grade A), SW-centered University Milestone Scholarship (merit-based)Pusan National University · 2020
Korean (native)English (professional working proficiency)
Teaching and Mentoring
Research Mentor - KUDoc, Korea University
2024 - Present | Seoul, Republic of Korea
Mentor 3 M.S. students and 2 Ph.D. students on document AI, multimodal RAG, and structured evidence reasoning.
Guided mentees’ research, resulting in 15 co-authored publications at top-tier venues.
Teaching Assistant & Research Mentor - Pusan National University
2019 - 2022 | Busan, Republic of Korea
Capstone research mentoring (2021–2022): Mentored undergraduate capstone teams on Korean dependency parsing and Table QA; one team achieved SOTA on the KLUE-DP benchmark.
PNU Software Hackathon (2021): As TA for Prof. Hyuk-Chul Kwon's Software Design course, mentored 100+ students across 30+ projects; seven student teams won awards including the Grand Prize.
Course teaching assistant: Software Design (Android, Java) and Web Application Programming under Prof. Hyuk-Chul Kwon; Programming Principles (C) under Prof. Eunmi Jeong.
