Wonjin Kim is a T-shaped AI researcher and engineer specializing in physics-informed deep learning, computational optics (OCT/CT), and LLM-based system design. His work bridges fundamental research and production engineering, with a focus on building research-to-production (R2P) AI systems that integrate medical imaging models, backend APIs, and LLM/RAG pipelines.

He is currently an AI Researcher at Nova Science, and concurrently a PhD Candidate (ABD) at KAIST.
This dual role is enabled by a strategic R&D collaboration between Nova Science and KAIST, allowing his research in computational imaging to directly inform the development of next-generation OCT/film technologies and commercial AI-assisted imaging products.

Previously, he conducted advanced medical imaging research at the Ewha AIBI Lab, served as a Technical Lead in healthcare platform development, and gained foundational experience in enterprise-grade system stability through his work with DB2 mission-critical systems at IBM Korea.

Wonjin has authored peer-reviewed papers, holds multiple patents in AI-powered medical imaging, and designs AI systems that unify theoretical rigor with real-world deployment. His personal blog documents work across AI research, imaging systems, and long-form conceptual synthesis using LLM reasoning tools.


Expertise / Skill Areas

  • Deep Learning / Computer Vision: PyTorch, Unsupervised Learning, Physic-informed deep learning
  • Biomedical Imaging: IV-OCT, Low-Dose CT, X-ray Fluoroscopy
  • AI System Architecture: Research-to-Production workflows, API Design
  • LLM Systems & RAG: Prompt Engineering, Reasoning Pipelines, Interdisciplinary Concept Synthesis
  • Backend / Enterprise Engineering: Python, TypeScript/JavaScript, DB2 Systems, Enterprise Technical Support

Wonjin actively investigates how LLM reasoning, hierarchical retrieval, and structured memory can be combined to synthesize fragmented interdisciplinary knowledge.
This work has led to the development of AI-augmented analytical pipelines for modeling complex behavioral and psychological phenomena, forming the basis for ongoing theoretical and methodological publications.


Full Publications


Patents

  • A learning and restoration method for reducing image noise using a neural network
    신경망을 이용하여 영상의 노이즈를 저감하기 위한 학습 및 복원 방법 및 이를 수행하는 컴퓨팅 장치

  • Method for predicting vascular characteristics using multi-loss AI models
    서로 다른 손실 함수를 사용한 인공지능 모델을 기반으로 혈관 특징을 예측하는 방법 및 그 시스템

  • Method for predicting fractional flow reserve (FFR) using vascular OCT images
    혈관 OCT 영상을 기반으로 분획혈류예비력(FFR)을 예측하는 방법 및 그 시스템

  • Method for detecting vascular bifurcation in OCT images
    혈관 OCT 영상을 기반으로 분기되는 혈관을 탐지하는 방법 및 그 시스템

  • Method for removing medical image artifacts using artificial intelligence
    인공지능을 이용하여 의료 영상의 아티팩트를 제거하는 방법 및 그 시스템

  • Method for improving medical image quality using artificial intelligence
    인공지능을 이용하여 의료 영상의 품질을 제고하는 방법 및 그 시스템