Summary
Rohith Peddi is a PhD candidate and Graduate Research Assistant at UT Dallas with nine years of industry and research experience building neuro-symbolic, multi-modal systems for 3D scene understanding and robot imitation learning. He designs perception architectures and real-time task-guidance systems that fuse vision, language, and audio with probabilistic reasoning, and has driven diffusion-based and test-time adaptation techniques to improve robustness under distribution shifts. His work on relational anticipation, long-tail learning, and monocular-to-4D generative models has led to high-profile publications including CVPR (Highlight), ECCV (Oral), NeurIPS, and IROS (Oral). Prior to research, he built scalable backend platforms and real-time services at startups, giving him a rare combination of production engineering and cutting-edge ML research skills. Based in Dallas, he brings practical systems experience (PyTorch, CUDA, ROS, Isaac Gym) to foundational videoโlanguage and graph-based dynamic scene models. A subtle throughline in his profile is translating human interaction priors into robot-specific skill transfer, bridging interpretability and actionable robotic behavior.
9 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at The University of Texas at Dallas
Dual Degree [B.Tech + M.Tech], Ocean Engineering, Dual Degree [B.Tech + M.Tech], Ocean Engineering at Indian Institute of Technology, Kharagpur
Telugu, Hindi