RESEARCH CV
Janghoon Cho, Ph.D.
Staff Engineer · Qualcomm AI Research, Korea
janghoon.cho86@gmail.com · Google Scholar ↗
Research profile
I develop methods for efficient multimodal AI, focusing on visual token compression, long-video understanding, and multimodal retrieval. My goal is to help make multimodal AI run efficiently, both in the cloud and on devices with limited computational resources. My background spans submodular optimization, audio and speech processing, and speaker verification development toward deployment on mobile devices.
Professional experience
Qualcomm AI Research Korea
Staff Engineer · May 2018 – Present
- Research in efficient visual token compression and multimodal learning.
- Led speaker verification development, including model architecture improvements and multi-GPU training pipelines, in collaboration with commercialization teams.
- Research in multi-domain imbalanced learning, acoustic scene classification, and audio-motion sensor fusion.
SK hynix
Senior Engineer · August 2017 – May 2018
Solution Algorithm Team, NAND Flash Memory Group.
Education
Ph.D., Electrical Engineering — KAIST, February 2018
Dissertation: Summarizing Distribution: Submodular Probability Density Cover
Advisor: Chang D. Yoo
M.S., Electrical Engineering — KAIST, August 2011
Thesis: Underdetermined Convolutive BSS Based on a Three-stage Approach
Advisor: Chang D. Yoo
B.S., Electrical Engineering — KAIST, 2009
Publications
- Janghoon Cho, Jungsoo Lee, Munawar Hayat, Kyuwoong Hwang, Fatih Porikli, Sungha Choi. FLoC: Facility Location-Based Efficient Visual Token Compression for Long Video Understanding. ICLR 2026.
- Hyojin Park, Yi Li, Janghoon Cho, Sungha Choi, Jungsoo Lee, Taotao Jing, Shuai Zhang, Munawar Hayat, Dashan Gao, Ning Bi, Fatih Porikli. ForeSea: AI Forensic Search with Multi-modal Queries for Video Surveillance. ECCV 2026.
- Jungsoo Lee, Janghoon Cho, Hyojin Park, Munawar Hayat, Kyuwoong Hwang, Fatih Porikli, Sungha Choi. Generalized Contrastive Learning for Universal Multimodal Retrieval. NeurIPS 2025.
- Janghoon Cho, Sunghyun Park, Hyunsin Park, Hyoungwoo Park, Seunghan Yang, Sungrack Yun. Balanced Learning for Multi-Domain Long-Tailed Speaker Recognition. ICASSP 2024.
- Seunghan Yang, Debasmit Das, Janghoon Cho, Hyunsin Park, Sungrack Yun. Domain Agnostic Few-shot Learning for Speaker Verification. INTERSPEECH 2022.
- Simyung Chang, Hyunsin Park, Janghoon Cho, Hyoungwoo Park, Sungrack Yun, Kyuwoong Hwang. Subspectral Normalization for Neural Audio Data Processing. ICASSP 2021.
- Seungwoo Yoo, Hee Seok Lee, Heesoo Myeong, Sungrack Yun, Hyoungwoo Park, Janghoon Cho, Duck Hoon Kim. End-to-End Lane Marker Detection via Row-wise Classification. CVPR Workshops 2020.
- Sungrack Yun, Janghoon Cho, Jungyun Eum, Wonil Chang, Kyuwoong Hwang. An End-to-End Text-Independent Speaker Verification Framework with a Keyword Adversarial Network. INTERSPEECH 2019.
- Janghoon Cho, Sungrack Yun, Hyoungwoo Park, Jungyun Eum, Kyuwoong Hwang. Acoustic Scene Classification Based on a Large-Margin Factorized CNN. DCASE 2019.
- Hyoungwoo Park, Sungrack Yun, Jungyun Eum, Janghoon Cho, Kyuwoong Hwang. Weakly Labeled Sound Event Detection Using Tri-training and Adversarial Learning. DCASE 2019.
- Janghoon Cho, Chang D. Yoo. MAP-based Permutation Alignment for Underdetermined Convolutive Blind Source Separation. IEEE ICCE-Asia 2016.
- Janghoon Cho, Chang D. Yoo. Underdetermined Convolutive BSS: Bayes Risk Minimization Based on a Mixture of Super-Gaussian Posterior Approximation. IEEE/ACM TASLP · 23(5), 828–839, 2015.
- Janghoon Cho, Chang D. Yoo. A Maximum Likelihood Approach for Underdetermined TDOA Estimation. ICASSP 2013.
- Janghoon Cho, Hyunsin Park, Chang D. Yoo. Blind Speech Separation and Recognition System for Human Robot Interaction in Reverberant Environment. URAI 2012.
- Janghoon Cho, Jinho Choi, Chang D. Yoo. Underdetermined Convolutive BSS Based on a Novel Mixing Matrix Estimation and MMSE Based Source Separation. IEEE MLSP 2011.
Selected research projects
- Speaker verification algorithms for mobile devices · 2020–2023
- Acoustic scene classification · 2019–2020
- Car-entry detection using audio-motion sensor fusion · 2018–2019
- Submodular active learning for lifelong machine learning · 2014–2017
- Acoustic user interfaces in vehicle environments · 2013–2014
- Satellite image denoising · 2011–2012
- Sound source localization and separation for human–robot interaction · 2009–2013
Recognition and technical background
Presidential Science Scholarship, Korea · 2004
Programming: Python, C/C++, MATLAB, Java.