About Me
I am building intelligence at Amazon AGI, as an Applied Scientist on Nova vision skill team.
I have been working on multi-modal/omni LLMs (Nova2 family) and multimodal embedding model (Nova MME), with a focus on multi-modal data. My specific interests lie in video(+audio), multi-modal agents, and data-centric AI.
I received my Ph.D degree from the Department of Computer Science and Engineering at Michigan State University in 2024,
advised by Prof. Yu Kong.
During Ph.D, I worked on video+language and human action detection. My thesis is Action Modeling in Long-form Videos. I have interned at AWS AI lab and Microsoft Azure.
Contact: junwenchen dot swjtu at gmail dot com
Research Interests
General interests LLM, Multi-modal Understanding, Data-centric AI.
Specific interests Omni Models (Video/Audio), Computer Use, Tool Calling, Video Editing, Streaming Understanding, Speech2Speech, Trustworthy Computer Vision.
Model Launch
Nova 2 Family.
Multimodal reasoning and generation models.
December 2025
Nova Multimodal Embedding.
State-of-the-art multimodal embeddings for agentic RAG and semantic search.
October 2025
Publications
ATM: Action Temporality Modeling for Video Question Answering.
Junwen Chen, Jie Zhu, Yu Kong
ACM Multimedia
Ottawa, ON, Canada. October, 2023
[ local copy | code ]
GateHUB: Gated History Unit with Background Suppression for Online Action Detection.
Junwen Chen, Gaurav Mittal, Ye Yu, Yu Kong, Mei Chen
CVPR
New Orleans, LA, USA. June, 2023
[ local copy ]
Explainable Video Entailment with Visually Grounded Evidence.
Junwen Chen, Yu Kong
ICCV
Online, October, 2021
[ local copy ]
Group Activity Prediction with Sequential Relational Anticipation Model.
Junwen Chen, Wentao Bao, Yu Kong
ECCV
Online, August, 2020
[ code ]
Activity-driven Weakly-Supervised Spatio-Temporal Grounding from Untrimmed Videos.
Junwen Chen, Wentao Bao, Yu Kong
ACM Multimedia
Online, October, 2020
RIT-18: A Novel Dataset for Compositional Group Activity Understanding.
Junwen Chen, Haiting Hao, Hanbing Hong, Yu Kong
Women in Computing Workshop(WiCV), CVPR
Online, June, 2020
[ dataset ]
Automatic Defect Detection of Fasteners on the Catenary Support Device Using Deep Convolutional Neural Network.
Junwen Chen, Zhigang Liu, Hongrui Wang, Alfredo Núñez, Zhiwei Han
IEEE Transaction on Instrumentation and Measurement
2018
Intelligent condition monitoring of railway catenary systems: A Bayesian network approach.
Hongrui Wang, Alfredo Núñez, Rolf Dollevoet, Zhigang Liu, Junwen Chen
Symp. Dyn. Vehicles Roads Tracks (IAVSD)
2017
Professional Service
- Program Committee: IJCAI 2023
- Reviewer: CVPR 2021-present, ECCV, ICCV, Multimedia, AAAI