Joongmin Shin

I research structure-aware Multimodal Reasoning: methods that recover document structure (layout, tables, figures, hierarchy, cross-page dependencies, and provenance) to turn complex documents into auditable evidence for reliable retrieval, grounded generation, and long-context reasoning. My first- and co-first-author work spans hierarchical retrieval, multimodal dependency parsing, structure-preserving chunking, and domain-specific RAG. My current manuscripts extend this agenda toward evidence auditing, multimodal QA evaluation, and evidence-centric memory for reliable multimodal agents.

I completed my B.S. in Computer Science and Engineering (advised by Prof. Jeongu Kim) and M.S. in Artificial Intelligence (advised by Prof. Hyuk-Chul Kwon) at Pusan National University, beginning NLP research in Prof. Kwon's AI Research Lab in 2020. I then joined Korea University's Human-Inspired AI Research Group, co-advised by Prof. Jaehyung Seo and Prof. Heuiseok Lim, where I founded and lead KUDoc, earning a promotion to Senior Researcher.

Together we have produced 13 publications and 13 manuscripts under review at top-tier venues, including 4 first-author papers at ACL, CVPR, and EMNLP, along with 5 patents, 4 industry projects with 3 technology transfers, and 7 awards.

I am seeking Ph.D. opportunities with advisors working on multimodal reasoning, structured memory, world models, and planning for future agents. If these directions resonate with yours, please feel free to reach out via email or LinkedIn.