Jihwan Eom ‍ Email & Phone Number

대한민국 경기도

Applied ML Dev TeamDeveloped a multilingual (EN/KO/JP/TH/TW) on-device vector search model from English-only pre-trained weights. Productized a non-autoregressive, on-device image captioning model for real-time messaging services.

대한민국

AI CAD(Computer Aided Diagnosis) Research TeamResearched an image-to-text model for generating radiology report from Chest-Xray (CXR) images.Addressed complex real-world challenges, focusing on noisy and imbalanced datasets.Pioneered in-domain weight initialization methods for CXR, including DINOv2, MAE, CoCa, etc.

대한민국 서울

대한민국 서울

Productized the Intel AI platform, Geti, as a training algorithm engineer.Researched and developed semantic segmentation, multi-label/head classification models in Geti.Proposed label-efficient, semi-supervised learning methods to enhance semantic segmentation.

연세대학교 의과대학 세브란스병원

Established a machine learning pipeline for radiomics research.Refined a CycleGAN model to reduce image bias in MRIs from filming devices of multiple centers.

Dependable Computing Lab.

Developed a deep learning model to detect stress using time-series data from wearable devices.

대한민국 일산

Built a model to predict prognosis scores from CXRs for the accurate intensive care unit allocation.