Jai Bardhan
Jai Bardhan is a researcher at the Czech Institute of Informatics, Robotics and Cybernetics (CIIRC) at the Czech Technical University in Prague, where he works on robot world models with Josef Sivic and Vladimir Petrik. He is the lead author of DepthWorld, a Stable Video Diffusion-based 3D world model for robot manipulation, and PersistWorld, which stabilizes multi-step world model rollouts via reinforcement learning, both published at ECCV 2026. He previously served as a predoctoral research fellow at TCS Research in the Deep Learning and AI Lab and completed his master's at IIIT Hyderabad, where he received the Best Thesis Award and Gold Medalist Award for the CND Department.
“DROID-3D—a calibrated 3D corpus providing dense metric depth and recalibrated multi-view extrinsics for over 70,000 episodes”
Source→“DepthWorld, a Stable Video Diffusion-based world model that jointly predicts multi-view RGB and depth via spatial latent tiling”
Source→“At equal training budget, DepthWorld gains +1.48 dB PSNR over an RGB-only baseline of identical architecture and data”
Source→“23× improvement on EE, 13× on WE, 8× on RD, and +37 absolute points on IoU”
Source→“also an author of PersistWorld (reference [2], RL-stabilized world models) and REALM (reference [42])”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.