Donghwa Kim

I'm a software engineer at Samsung Research. I got my M.S degree from Graduate School of Data Science (GSDS) at Seoul National University. I was advised by Professor Joonseok Lee, as a member of Visual Information Processing Lab (VIPLab). Prior to my graduate studies, I majored in astronomy and minored in physics.

Email  /  CV  /  Scholar  /  Github

Research

I'm interested in computer vision, multimodal representation learning, foundation models. Most of my research is about analyzing existing mulimodal model and improving its application capabilities.

We propose a simple but powerful data augmentation method which augments a training image into a mosaic with three other negative images carefully curated by a pretrained multimodal alignment model, e.g., CLIP, to make the sample more challenging.

We conduct an extensive benchmark study to measure the performance of representative methods on widely used 7 datasets, while posing additional research questions and empirically verify them.

Work Experience

Oct. 2024 – Present

Seoul, Korea

Samsung Electronics  ·  Software Engineer

Samsung Research → AX Team (Management Diagnosis Office)

DoXA — Document Extraction and Analysis  (Internal Service, 2025.01–2025.12)

• Developed a document QA benchmark dataset targeting image-heavy document understanding, supporting evaluation of multimodal parsing capabilities
• Owned end-to-end production deployment of internal AI service, including K8S setup, PostgreSQL HA configuration, and BE/FE development

Ehcro — Multi-Agent Orchestration Engine  (Internal Service, 2026.01–2026.05)

• Managed production deployment of LLM-based multi-agent service, establishing CI/CD pipelines and conducting cluster-level reliability testing

Rosetta — Internal Translation Agent Service  (2026.06–Present)

• Building an agentic translation service for product manuals across Mobile eXperience and Digital Appliances business, handling both BE and FE development

Sep. 2021 – Dec. 2021

Seoul, Korea

SETsystem  ·  AI Research Intern

• Surveyed SOTA models for object detection and semantic segmentation to detect ship & wave on radar maps
• Replaced Mask R-CNN with a simpler U-Net architecture, improving mIoU by over 1%

This page is a fork of Jon Barron's. Thank you for sharing :)