Taehoon Kim

Taehoon Kim

Integrated M.S./Ph.D. Student, MIIL, Korea University · Seoul, South Korea

About

I am an integrated M.S./Ph.D. student at the Multimodal Interactive Intelligence Laboratory (MIIL), Korea University, which I joined in September 2025. My current work is on multi-modal large language models and structured visual generation — most recently, getting VLMs to produce SVGs that are hierarchically organized and genuinely editable rather than flat collections of paths.

Before returning to research full time, I spent eight years at Samsung SDS, where I led the end-to-end development of AICR, a SaaS-based OCR service — from architecture and model research through deployment. That work still shapes what I care about: document AI, table detection and structure recognition, and OCR.

Research interests: Multi-modal Large Language Models, Vision-Language Models, Structured & Compositional Generation, Computer Vision, Document AI, OCR, Table Extraction.

Publications

* denotes equal contribution.

Conference

Compositional SVG Generation via VLM-Driven Hierarchical Semantic ParsingMain

Sehwan Park*, Taehoon Kim*, Geonhee Han, Dohyun Kim, Seung Wook Kim & Paul Hongsuck Seo

Conference on Empirical Methods in Natural Language Processing (EMNLP), 2026.

arXivProject page

Training method that integrates Table Detection and Table Structure Recognition datasets based on a Multi-Task learning deep learning model

Taehoon Kim

37th Workshop on Image Processing and Image Understanding (IPIU), 2025.

End-to-End Table TransformerOral

Yun Young Choi*, Taehoon Kim*, Namwook Kim, Taehee Lee & Seongho Joe

International Conference on Document Analysis and Recognition (ICDAR), 2024.

DOI

Journal

Scheduling appointments for container truck arrivals considering their effects on congestion

Sanghyuk Yi, Bernd Scholz-Reiter, Taehoon Kim, Kap Hwan Kim

Springer Science+Business Media, LLC, part of Springer Nature, 2019.

Patents

Experience

Researcher · Samsung SDS

Jan. 2018 — Aug. 2026

Seoul, South Korea

  • Led the end-to-end development of AICR, a SaaS-based OCR service:
    • Designed the architecture for scalability and efficiency.
    • Researched and developed improvements to OCR accuracy and performance.
    • Managed service deployment and distribution.
  • Contributed to Samsung Nexfinance AICR, a deep-learning-based OCR solution (v1.1 – v1.3).
  • Implemented Kyongnam Bank's credit and collateral evaluation system.
  • Developed the search engine for Samsung SDS Brity Assistant, a corporate chatbot system.

Assistant Researcher · Logistics Innovation Networking Labs

Jul. 2016 — Dec. 2017

Busan, South Korea

  • Participated in developing a vehicle reservation system on an IoT-based collaboration platform.
  • Designed and implemented core algorithms for vehicle reservation efficiency.

Education

Integrated M.S./Ph.D. Program

Sep. 2025 — Present

Multimodal Interactive Intelligence Laboratory (MIIL), Korea University, Seoul, South Korea

M.S. in SW/AI Convergence

Mar. 2023 — Aug. 2025

Korea University, Seoul, South Korea

B.S. in Computer Science and Engineering

Mar. 2011 — Feb. 2018

Pusan National University, Busan, South Korea

Awards

Skills

Programming
Python, C++, C, Java, JavaScript, Go
Frameworks & Tools
PyTorch, OpenCV, Docker, Kubernetes, Linux (Ubuntu, CentOS)