Krish Pandya

Hi, I'm Krish, an undergraduate researcher at the Robotics Research Center, IIIT Hyderabad, advised by K. Madhava Krishna and working with Sourav Garg. I'm interested in how a robot should store the world it moves through so it can navigate, which I chase through SLAM, 3D reconstruction, and topological maps.

I also spend a lot of time in open source. I am on the Sugar Labs board, write a fair amount of Rust, and build reproducible setups with Nix.

When I'm not working, I make and analyze music, read philosophy, play sports and video games, and watch a ton of stuff. I also like reading about how perception and learning work in brains and machines.

Reach me at bruvistrue93 at gmail dot com if you'd like to talk about research, collaboration, open source, or life.

Publications

Plann3r: predicting planning costs grounded in 3D

Aditya Vadali*, Krish Pandya*, Vansh Garg*, Nishant Chandna, Dustin Craggs, Feras Dayoub, K. Madhava Krishna, Sourav Garg

* Equal contribution.

Predicts pixel-level planning costs as geodesic distances from a query image to an arbitrary subgoal pixel, using a VGGT backbone with a learnable goal token and an MLP cost decoder. The VGGT-Nav pipeline runs the same module twice: offline for mapping and global planning, online for localization and local planning that conditions a learnt controller.

News

Experience

Teaching

Talks and demos

Writing

The rest is on Substack.