The University of Texas at Austin

Indian Institute of Technology, Kanpur

IIT Kanpur

Ph.D. Candidate (Robotics)  -  Robot Learning  -  Robot Manipulation  -  Motion Planning  -  AI, ML  -  Deep Learning  -  Controls

- POMDP-HD motion planner that generates efficient plans (via trajectory optimization) leveraging object interactions to perform long-horizon manipulation tasks with high accuracy even under uncertainty- A novel method for learning planning-compatible hybrid kinematics models for articulated objects from human demonstrations (improved up to 3x in accuracy over the state-of-the-art method)- Deep learning-based method to estimate articulation models for objects directly from raw depth images without knowing their articulation type a priori using screw theorySkills: Motion Planning, Model Learning, Deep Learning, Sim2Real

Graduate Research Assistant

- Developed a fully autonomous training pipeline to train networks for generating category-level optimal grasping behaviors given a single RGB-D image of the scene via imitation learning- Achieved a high success rate(∼70%) in grasping 40 different objects belonging to two object categoriesSkills: Deep Learning, Grasp Planning

Samsung AI Research Center, New York

Research Scientist - Robotics (Intern)

- Implemented state-of-the-art visual servoing algorithms to do high precision (≤2mm) assembly tasks. Ensured fast convergence (≤20s) to target features with high repeatability (≥90%success)- Developed python-based interfaces ensuring a task and platform-agnostic implementationSkills: Motion Planning, Perception, Software Development

Vicarious AI

Roboticist

Led two classes of 20 students each to conduct laboratory experiments for modeling and control of physical systems and tutored them in control and data acquisition using LabVIEW.

Teaching Assistant

- Led a class of 105 students in ME Lab-II (ME 471N) to conduct manufacturing laboratory experiments on hydraulic die press machine- Tutored a class of 55 students in Optimization Methods in Engineering Design (ME 752) course

- Modeled dynamics of different holonomic and non-holonomic mobile robot platforms and implemented motion planning library FIRM to perform navigation tasks under uncertainty- Interfaced robotic simulator, V-Rep, with the library with added features including customizable work environments and synchronized communication for real-time controlSkills: Motion Planning, Software Development