Darshan Hegde

Current Research Interests: Post-training and fine-tuning of robotic behavior models. More specifically around reward models, offline RL and online RL methods to make them reliable (>99.9% success rate) and performant (matching human throughputs).

I also write commentaries and summaries on latest research in the field. Previously, worked on behavior models for tactile multi-fingured dexterity at Tacta Systems, Inc, worked on grasping at Dexterity, Inc, worked on perception & prediction for self-driving at Aurora Innovation & Uber ATG & NVIDIA.

I graduated from Center for Data Science, New York University with MS in data science with deep learning & high performance computing concentration. I was also a part-time student at Robotics Graduate Program, Stanford University.

Curriculum Vitae: here.