I'm a PhD candidate at Nanyang Technological University (School of Electrical & Electronic Engineering) and an A*STAR CIS Scholar at the Institute for Infocomm Research (I²R), advised by Prof. Xudong Jiang. I study multimodal video understanding, curious what a model can really afford to remember.
That curiosity shapes how I build: state-space models and spatiotemporal compression that let vision-language foundation models perceive and reason over long videos at tractable cost. This work has produced first-author papers at CVPR 2025 (TV3S), ECCV 2026 (STAC), and in Machine Intelligence Research. In late 2025 I spent a research visit with Prof. Yun Liu's group at Nankai University, Tianjin.
Before the PhD, I earned First-Class Honours at NTU while working full-time as a research assistant, leading R&D that secured over S$200K in government funding and mentoring 40+ students to national competition wins.
European Conference on Computer Vision, 2026 ECCV 2026
STAC compresses the spatiotemporal token stream feeding a multimodal LLM: roughly 85% fewer tokens and a 1.8× speedup in training and inference, while keeping reasoning-segmentation accuracy competitive.
IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025 CVPR 2025 22.1% Acc. Rate
TV3S shares Mamba state spaces across time so temporal context propagates through a selective gate instead of a memory-heavy feature pool, enabling parallel training and inference over long videos.
Machine Intelligence Research, accepted 2026 Journal
A systematic study of what SAM2's video foundation model does, and does not, bring to dense, class-aware video segmentation.
arXiv:2506.13552, 2025 Under Review
A unified review of video scene parsing across semantic, instance, panoptic, tracking, and open-vocabulary segmentation, tracing the field from convolutional pipelines to foundation-model approaches.
IEEE 19th Conference on Industrial Electronics and Applications (ICIEA)
IEEE 18th Conference on Industrial Electronics and Applications (ICIEA)
Journal of Automation and Intelligence
IEEE 17th Conference on Industrial Electronics and Applications (ICIEA)
IEEE 17th Conference on Industrial Electronics and Applications (ICIEA)
IEEE 17th Conference on Industrial Electronics and Applications (ICIEA)
IEEE 16th Conference on Industrial Electronics and Applications (ICIEA)
IEEE 16th Conference on Industrial Electronics and Applications (ICIEA)
IEEE International Conference on Multimedia & Expo Workshops (ICMEW)
* denotes equal contribution
→ View all publications on Google Scholar
Prof. Yun Liu's group, enhancing multimodal understanding in video models.
Topic: Effective Visual Perception with Large Foundation Models
Advisor: Prof. Xudong Jiang · Expected: Dec 2027
Full-time research role held concurrently with my part-time bachelor's degree.
FYP: Real-time SLAM using deep features
GPA: >4.9/5.0 First-Class Honours Dean's List ×4FYP: Improving Recognition Performance for Low-Resolution Images Using DBPN
Director's Roll of Honour GPA: >3.9/4.0A*STAR I²R, Fusionopolis
Rapid-Rich Object Search (ROSE) Lab, NTU Singapore
