Jai Bardhan

Hello 👋 !

My name is Jai.

I am a researcher in the IMPACT Group at CIIRC, Czech Technical University in Prague, advised by Josef Sivic and Vladimir Petrik. I work on world models for robotics: 3D-consistent action-conditioned video models, world action models, RL post-training of world models, and using them to evaluate and improve robot policies. More broadly, I am interested in generative and self-supervised learning.

Recent projects:

• DepthWorld (CoRL 2026): a 3D world model that jointly predicts multi-view RGB and metric depth, trained on DROID-3D, 71k DROID episodes we recalibrated to 4.9 mm median extrinsic error.
• PersistWorld (ECCV 2026, 🏆 Outstanding Paper Award at UniWorld): RL post-training that keeps video world models stable over long autoregressive rollouts.
• CSCD (WACV 2026): curve skeletonization of meshes and point clouds in the continuous domain, via discrete differential geometry.

Before CIIRC, I spent two years as a predoctoral research fellow in the Deep Learning and AI Lab at TCS Research with Ramya Hebbalaguppe, working on 3D shape analysis, confidence calibration in vision-language models, and evaluating and interpreting diffusion models.

I did my B.Tech. in Computer Science and M.S. by Research in Computational Natural Sciences at IIIT Hyderabad. My M.S. thesis with Subhadip Mitra, “Revisiting Deep Learning for Particle Physics”, received the Ritesh Tiwari Outstanding MS Thesis Award. I still collaborate with Subhadip’s group on machine learning for particle physics, such as self-supervised foundation models for collider data (HEP-JEPA).

I am always happy to chat about world models and robot learning, so feel free to reach out!

You can find my CV here.

awards & service

• 2026 · Outstanding Paper Award & Oral, UniWorld Workshop @ ECCV 2026, for PersistWorld
• 2026 · Gold Reviewer, ICML 2026
• 2023 · Ritesh Tiwari Outstanding MS Thesis Award, IIIT Hyderabad
• 2023 · Gold Medal (highest CGPA, Computational Natural Sciences, 2018 batch), IIIT Hyderabad
• Reviewer · CoRL, ECCV, CVPR, SIGGRAPH Asia, BMVC, NeurIPS, ICML, AAAI, TMLR