Trung Pham

Principal Research Scientist · NVIDIA — Cosmos Lab · Santa Clara, CA

trungp@nvidia.com·trung.ptt@gmail.com
trungtpham.github.io·Scholar·GitHub·LinkedIn

Principal Research Scientist at NVIDIA Cosmos Lab, building world foundation models for Physical AI. Track record across research, product delivery, and technical leadership. Ph.D. in Computer Vision; Google PhD Fellow and Heidelberg Laureate Forum delegate.

Experience

Principal Research Scientist — NVIDIA
2018 — Present

Build and scale world foundation models for Physical AI — multimodal systems that perceive, simulate, and reason about the physical world.

Technical lead for AV perception — architected and shipped production BEV and camera perception models into NVIDIA's autonomous driving stack.

Postdoctoral Research Fellow — Australian Centre for Robotic Vision, University of Adelaide
2014 — 2018

Research on 3D scene understanding and instance segmentation. Core team member on the 1st-place entry at the 2017 Amazon Robotics Challenge.

Software Engineering Intern — Google
Aug — Oct 2013

Efficient placement algorithms for a global storage system.

Education

Ph.D., Computer Vision — The University of Adelaide
2010 — 2014

Computer Vision, Machine Learning. Google PhD Fellow. Dean's Commendation for Doctoral Thesis Excellence.

M.Eng., Computer Engineering — Chonnam National University
2008 — 2010

Brain Korea 21 Scholar.

B.Sc., Mathematics and Computer Science — Vietnam National University, HCMC
 

Honors

Skills

Expertise World foundation models · VLM · VLA · video generation & simulation · radar simulation · BEV perception · online mapping · end-to-end autonomous driving · 3D scene understanding · semantic segmentation · object detection · motion estimation · robust estimation. Stack Python · PyTorch · C++ · SLURM · large-scale distributed training. Leadership Technical lead · mentoring scientists and PhD interns · cross-functional teams.

Selected Publications

Cosmos 3: Omnimodal World Models for Physical AI 2026 NVIDIA. NVIDIA Technical Report.
World Simulation with Video Foundation Models for Physical AI 2025 NVIDIA. arXiv:2511.00062.
NVAutoNet: Fast and Accurate 360° 3D Visual Perception for Self-Driving 2024 T. Pham, M. Maghoumi, et al. WACV 2024.
Bayesian Semantic Instance Segmentation in Open Set World 2018 T. Pham, V. B. G. Kumar, T.-T. Do, G. Carneiro, I. Reid. ECCV 2018.
Cartman: The Low-Cost Cartesian Manipulator that Won the Amazon Robotics Challenge 2018 D. Morrison, …, T. Pham, et al. ICRA 2018.
A Bayesian Data Augmentation Approach for Learning Deep Models 2017 T. Tran, T. Pham, G. Carneiro, L. Palmer, I. Reid. NeurIPS 2017.
Meaningful Maps — Object-Oriented Semantic Mapping 2017 N. Sünderhauf, T. Pham, Y. Latif, M. Milford, I. Reid. IROS 2017.
Efficient Point Process Inference for Large-Scale Object Detection 2016 T. Pham, H. Rezatofighi, T.-J. Chin, I. Reid. CVPR 2016.
The Random Cluster Model for Robust Geometric Fitting 2013 T. Pham, T.-J. Chin, J. Yu, D. Suter. IEEE TPAMI 2013.

Service

Reviewer for top-tier computer vision, machine learning, and robotics venues — CVPR, ICCV, ECCV, NeurIPS, ICRA, IROS, IEEE TPAMI, and IEEE TIP.