Allen Tu

I am a second-year Computer Science Ph.D. student at the University of Maryland, College Park, advised by Professors Tom Goldstein and Matthias Zwicker. My research focuses on efficient and reliable 3D/4D reconstruction and generation, including compression methods for Gaussian Splatting and diffusion-based priors that enable reconstruction from sparse or degraded imagery. I also worked on multimodal biometric recognition under severe operating conditions. I am currently an Applied Science Intern at Amazon, building photorealistic simulation environments for training embodied AI by reconstructing physics-ready 3D digital twins from real-world sensor data.

My research is supported through the IARPA WRIVA and BRIAR programs and shaped by collaborations across academia and industry, including prior internships at Systems & Technology Research (STR) and my current role at Amazon. I co-organized the inaugural SPAR-3D Workshop at CVPR 2026 on security and adversarial robustness in 3D generative vision models, and previously mentored undergraduate researchers in the FIRE: Capital One Machine Learning program. I received my B.S./M.S. in Computer Science from UMD in 2024.