Ph.D. Candidate in Robotics ยท University of Michigan
I am a member of the ROAHM Lab, advised by Prof. Ram Vasudevan. I work on the perception, modeling, and planning of deformable objects. My dream demo is a fully autonomous wire harness assembly.
Previously, I worked in the MMint Lab with Prof. Nima Fazeli on visuo-tactile representation learning for robotic manipulation.
* equal contribution
IJRR 2026 The International Journal of Robotics Research (accepted)
A differentiable physics model with residual learning for branched cables: momentum-conserving junction constraints, real-time inference above 100 Hz, and planning for shape matching and thread insertion.
IROS 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems
Recovers 3D keypoints and their topology from RGB-D recordings of cables, branched cables, fabric and cloth, and tracks them without markers or object-specific training.
CoRL 2024 Conference on Robot Learning
DEFORM combines differentiable discrete elastic rods with learning to model ropes and cables accurately and in real time, and uses the model to track a cable through occlusions.
CoRL 2022 Conference on Robot Learning
A multimodal transformer for model-based reinforcement learning that uses tactile feedback with self- and cross-modal attention to focus on task-relevant regions of the image.