Keondo Park

PhD student

A2I group, Graduate School of Data Science, Seoul National University

Office: 3rd floor, Graduate school of data science, building 942, SNU, Seoul, South Korea

Contact: gundo0102@snu.ac.kr

github: KeondoPark (github.com)

Ph.D in Data Science, Seoul National University, 2022-

M.S. in Data Science, Seoul National University, 2020-2022

B.S. in Math, Seoul National University, 2007 - 2012

Milliman Seoul office, 2013 - 2020

Milliman UK office, 2016

Associate of Society of Actuaries(ASA)

Chartered Financial Analyst(CFA)

[IPSN] PointSplit: Towards On-device 3D Object Detection with Heterogenous Low-power Accelerators, 2023, Keondo Park, Yourim Choi, Inhoe Lee, Hyungsin, Kim.
(Video)

[arxiv] Federated Semi-Supervised Learning with Prototypical Networks, 2022, Woojung Kim, Keondo Park, Kihyuk Sohn, Raphael Shu, Hyung-Sin Kim.

Best Paper Award from 2022 summer Korean AI Conference: Federated Semi-Supervised Learning with Prototypical Networks

Best Paper Award from 2022 Korean Sleep Medicine Conference: On-device Real-time Sleep Stage Classification with Single Channel EEG using Low-power NPU

2022 BK outstanding graduate student 

2021 GSDS InnoJam winner

2nd place in 2022 national AI competition (semi-supervised image segmentation).

2nd place in 2022 Naver AI Rush (landmark detection)

4th palce in 2021 national AI competition (lightweight object detection)

Excellence award in 2022 k-ium medical AI competition

I am interested in developing lightweight neural network models suitable for mobile devices including Google Coral, Jetson Nano, smartphones or IoT. This includes reducing large model sizes to fit in small devices. Pruning, quantization, or knowledge distillations are popular example techniques for this purpose.

I am collaborating with Hyunwoo Shin from SNU college of medicine to develop lightweight sleep staging algorithms from single EEG that could be readily deployable to small edge devices. I believe this would enhance the diagnosis and treatment of sleep-related disorders.

I like to utilize new gadgets(e.g. Coral) in our everyday life. For example, Coral could be used to detect whether people are wearing masks or not. It was introduced in a newspaper: https://www.mk.co.kr/news/it/view/2020/06/669932/  

Participating in SNU Project Group for Education and Research in Medical AI

Class Projects

• 2023  Ambient AI : On-device Real-time Sleep Stage Classification with Single Channel EEG on Coral (video)

•  2020 ML/DL:  Mask Detection with Google Coral Board (video)