Contact [email protected]

Github seungwoos

Linkedin Seungwoo Kim

I’m a machine learning engineer interested in making our lives beneficial and productive. I had worked in medical imaging and computer vision, and currently digging on Large Language Models, but not limited to.

As of December 2023, I am working as an AI Engineer at 3billion. Previously, I worked as a (front-end) software engineer at Dancers Connect.

Experiences

AI Engineer @ 3Billion

Dec. 2023 - Present

• Working on large language models (LLMs) for diagnosis.
  • From instruction dataset generation to fine-tuning PEFT-based open-source LLM model.
  • Building Retrieval Augmented Generation (RAG) using knowledge using dify.

Software Engineer @ Dancers Connect

Oct. 2023 - Dec. 2023

• Worked as a front-end engineer, developed and maintained services at danceapply.com using ReactJS and Typescript.

Research Assistant @ UNIST

Mar. 2021 - Jun. 2023

• Implemented Neural Diffusion Processes architecture in PyTorch before the official code was released. [Code]
• Planned and conducted 2D image experiments using the proposed diffusion model in distributed setting. [Code]
  • Submitted to Conference, codes and papers will be released soon.
• Conducted experiments on 2D cytology and CT image generation using proposed diffusion model with acceleration.

Research Intern @ Tomocube AI Team

Mar. 2021 - Jul. 2022

• Conducted experiments on thyroid malignancy classification and proposed the combination method of bright field and refractive index image. [Code]
  • Submitted to Scientific Reports
• Implemented post-stenting stent prediction of Intravascular Ultrasound (IVUS) cardiovascular model based on vision transformer, and conducted experiments using different modalities (image and clinical information)

Collaborative Research with YUHS

Jan. 2022 - Jul. 2022

• Conducted experiments on brain malignancy differentiation of 2D CT image from YUHS using Deep Ensembles and Neural Bootstrapper
  • Submitted to Journal, codes and papers will be released soon.
• Conducted experiments on tumor segmentation of 3D Medical Segmentation Decathlon with synthetically generated imbalanced and noisy label datasets

Teaching Assistant @ UNIST

Sep. 2022 - Dec. 2022

• [AI502] Principles of Deep Learning (Instructor: Sungbin Lim)