- Master of Science in Robotics, WPI, 2020 (expected)
- Bachelor of Engineering in Mechanical Engineering, BITS Pilani, India (2018)
- Programming: Python, C/C++, MATLAB
- Deep Learning frameworks: PyTorch, Tensorflow, Keras
- Simulation softwares: Gazebo, OpenRave, OpenAI gym, Ansys, SolidWorks
- Libraries: OpenCV, Point Cloud Library, scikit-learn
- Robotic Framework: Robot Operating System
- Courses: Deep Learning for Advanced Robot Perceptioin, Artificial Intelligence, Robot Dynamics, Robot Control
- Peer Learning Assistant, Physics Department, WPI (Aug ‘19 - Present)
- Conduct lab sessions for PH 1110 (General Physics : Mechanics) and PH 1120 (Electricity and Magnetism)
- Tutor Active Learning Physics Course (PH 1110).
- Teaching Python to Physics majors.
- Computer Vision Group, MathWorks, Natick, MA (May ‘19 - Aug ‘19)
- Formulated an innovative CV algorithm to improve accuracy of camera calibration parameters for Fisheye Cameras.
- The Checkerboard Detection algorithm designed for Fisheye Cameras had better true positive detection even for imagesfrom Pinhole and Stereo Cameras.
- Achieved better checkerboard detection precision (98 %) as compared to the existing technique (83 %).
- Centre for Artificial Intelligence and Robotics, Bangalore, India (Jan ‘18 - June ‘18)
- Developed a novel image processing algorithm for efficient road segmentation in unstructured environment.
- Generated costmap in ROS using pointcloud information from Velodyne LIDAR, Stereo Camera and Ultrasonic sensor.
- Achieved better segmentation accuracy (91 %) as compared to existing Pyramid Scene Parsing Network (79 %)
- Learning based Motion Planning for Manipulators, WPI (Aug ‘19 - Present)
- Designing and applying DDPG to a 4 DOF manipulator to achieve motion planning faster than RRT.
- Adding expert demonstrations in the DDPG implementation to achieve faster convergence.
- Comparing and evaluating Imitation Learning, Supervised Learning and DDPG-MP approaches for motion planning.
- Viewpoint optimization for aiding grasp synthesis algorithms using supervised learning, WPI (Jan ‘19 - Present)
- Implemented active vision methodology to optimize depth sensor viewpoint to increase synthesized grasp quality.
- Employed supervised learning techniques to obtain viewpoint optimized policy by generating automated training data.
- Simulated results using Gazebo. Currently implementing the algorithms on Franka Emika Panda Robot.
- Ship Detection and Segmentation from Aerial Images, WPI (Aug ‘18 - Dec ‘18)
- Implemented a two model Deep Learning architecture to segment ships from aerial images on Airbus Dataset.
- Applied ResNet to classify images containing ships which were later fed to a stacked Hourglass model for segmentation.
- Control Lyapunov Barrier Strategy for Adaptive Cruise Control, WPI (Aug ‘18 - Dec ‘18)
- Combined Control Lyapunov Functions and Control Barrier Functions through Quadratic Programming to achieve Adaptive Cruise Control for a vehicle.
- Extended this control strategy in 2 Dimensions with incorporation of the dynamic model of turtlebot.
- Image Processing Techniques for Pothole Detection, BITS Pilani (Sept ‘16 - Dec ‘16)
- Implemented anomaly detection algorithm on pavement video data captured from a camera present on a quadcopter.
- Computed the location of the anomaly present on the pavement which was later transmitted to a ground robot for non- destructive testing.
Service and leadership
- Image Processing Lead and Treasurer of Radio Control Club, BITS Pilani.
- Contributed in organizing workshops as a student member for the American Society of Mechanical Engineers. (ASME).
- Conducted two day workshops, attended by 200+ college students, on ’Hovercraft building’ in March 2016 and ’Quadcopter demonstration’ in October 2016 at BITS, Pilani, India.