Junbum Cha

ML Research Scientist at TwelveLabs

Gwangmyeong-si, Gyeonggi, South Korea

About

As an AI research scientist and engineer, I am passionate about developing innovative core technologies that make a significant impact in the real world. My unique strengths lie in my ability to excel in both research and engineering, as well as my capacity for rapid learning, enabling me to solve complex problems across various domains. With a career that began as a software engineer, I transitioned through roles such as machine learning engineer and research engineer, finally leading to my current position as a research scientist. My diverse background equips me to excel in both research and engineering aspects, and though my recent accomplishments predominantly focus on research—resulting in 10 publications in the past 3 years, with five as the first author—my strong engineering skills are an essential contributor to my success as a research scientist. Throughout my career in the industry, I have gained experience across a wide range of research and application domains for different companies. I have worked on reinforcement learning for Go AI, scene-text recognition and generalization for OCR, image generation for font generation, segmentation, object detection, and vision-language modeling for multi-modal understanding. Currently, I am working on explainable AI to develop a trustworthy healthcare system. These successful transitions in diverse domains demonstrate my specialty in rapid learning and adaptability.

Experience

  • ML Research Scientist at TwelveLabs
    Jul 2025 - Present · 1 yr 1 mo

  • AI Research Scientist at Kakao Corp
    Jul 2024 - Jul 2025 · 1 yr 1 mo

    Led core development of Kanana-V, a production-ready Korean vision-language LLM with state-of-the-art performance, and contributed to Kanana-O, its omni-modal extension with audio I/O.

  • AI Research Scientist at 카카오브레인 - kakaobrain
    Jul 2021 - Jun 2024 · 3 yrs

    Unified foundation model team - Developing multimodal and multilingual agents, focusing on vision, language, audio modalities, and Korean-English bilingualism. Multimodal understanding team - Developed domain generalization algorithm for the large-scale era - Developed large-scale general visual language model for few-shot in-context learning - Developed unsupervised open-world semantic segmentation method - Developed open-vocabulary object detection method - Developed explanation method for the trustworthy healthcare system - Developed LLM-based healthcare radiologist agent

  • AI Research Engineer at NAVER Corp
    Mar 2019 - Jul 2021 · 2 yrs 5 mos

    OCR team - Developed automatic font generation systems - Developed robust training algorithm for distribution shift - Developed scene-text recognition engine - Published 7 papers, three as the first author

  • Machine Learning Engineer at NHN
    Mar 2017 - Mar 2019 · 2 yrs 1 mo

    - Developed superhuman-level AI go engine using reinforcement learning - Developed duplicated image search engine