CV

PhD Candidate

Kansas State University

PhD Candidate at Kansas State University. Research on agricultural application of Deep Learning and Robotics.

Positioning Engineer

Caterpillar

Developed on several Internal tools for efficient debugging and understanding of Positioning Systems, Drivers for IMU and GPS

Robotics Engineer

Rex Robotics

Worked on developing several controls algorithms for Quadrupedal Robots, Drones and Rovers. Worked on perception systems for the same.

R&D Robotics Engineer

Swarm Robotix

Worked on Dynamic modelling and Localization solutions to swerve drive robots

Graduate Research Assistant

Controls, Robotics and Mechatronics Lab

Focused on dynamic modelling and Controls of robotic arms.

PhD

Kansas State University

Biological and Agricultural Engineering

• Deep Learning
• Neural Networks in Engineering

Nanodegree

Udacity

Deep Learning

• Neural Networks
• Convolution Neural Networks
• Recurrent Neural Networks
• Deep Reinforcement Learning

Master

Cleveland State University

Mechanical Engineering

• Artificial Intelligence
• Robotics Dynamics and Control
• Human Motion Control
• Mechatronics

Bachelor

Jawaharlal Nehru Technological University

Mechanical Engineering

• Basic Electronics
• Kinematics
• Numerical Methods
• Dynamics
• Design of Machinery
• Fluid Mechanics
• Data Structures

Model Predictive Controller for a Quadruped

Designed a model predictive controller using QPOASES for a quadruped, involving linearization of a highly non-linear model, and achieving a running speed of 200Hz for robot stabilization.

• Quadratic Problem Solving
• Shooting Method
• QPOASES

Robust Passivity Controller with Regeneration

Developed and implemented a robust passivity-based controller for energy regeneration on a PUMA robot, using kinematics and Lagrange dynamic formulation, simulated in Simulink and Matlab.

• Kinematics and Lagrange Dynamic Formulation
• Simulink and Matlab Simulation
• D-space Implementation

Sliding Mode Controller with Regeneration

Built a refined model for an R-R-R-R manipulator and created a sliding mode controller, applied and tested through Dspace and Simulink simulations.

• Model Refinement
• Sliding Mode Control
• Dspace and Simulink

Optimization of Trajectory with Energy Regeneration

Applied inverse kinematics and optimized robot trajectory using a Non-linear solver, achieving a 20% efficiency increase in energy regeneration with IPOPT and Simulink.

• Inverse Kinematics
• Non-linear Optimization
• Energy Efficiency Improvement

Self-Playing Pac-man Game

Implemented AI algorithms to enhance the Pac-man agent's gameplay, achieving an improved average score over 1000 games through path planning and decision-making algorithms.

• AI Algorithms Application
• Path Planning and Decision Making
• Gameplay Improvement

Deep Deterministic Policy Gradient for Quadcopter Flight

Utilized DDPG to teach a quadcopter autonomous flight, designing actor and critic networks in Keras and achieving success in 134 episodes.

• DDPG
• Quadcopter Autonomous Flight
• Keras

Skid Steer Controller with Radius Constraints

Implemented a skid steer controller optimized to run at 100Hz on STM32 and Nvidia-Orin platforms, integrating microRos and Ros2 with constraints on velocity and turning radius.

• Skid Steer Control
• STM32 and Nvidia-Orin
• microRos and Ros2

MPCC Controller for Ackerman Car

Developed a Model Predictive Contouring Controller for an Ackerman vehicle, tested in Gazebo and ROS1 for potential agricultural applications.

• Model Predictive Contouring Control
• Gazebo and ROS1
• Agricultural Application Study

Ground Reaction Force Predictor

Developed a Python script using OpenCV for predicting ground reaction force from shoe curvature images, achieving an 80% accuracy with neural network training.

• Ground Reaction Force Prediction
• OpenCV
• Neural Network Training

Image Classifier Using Convolutional Neural Network

Created a CNN in TensorFlow by transferring features from Resnet-50, trained to identify dog breeds with an accuracy of 84.52% across 133 classes.

• CNN
• TensorFlow
• High Accuracy Classification

TV Script Generation with LSTM

Developed an LSTM network to generate TV scripts for The Simpsons, using season-long scripts for training and achieving content generation.

• LSTM Network
• Script Generation
• The Simpsons

Face Generation with GAN

Implemented a DCGAN to generate realistic human faces from random noise inputs, ensuring balanced network tuning.

• DCGAN
• Face Generation
• Network Tuning

Extended Kalman Filter for Sensor Fusion

Implemented an Extended Kalman Filter for fusing Lidar and Radar data, achieving an RMSE of less than 1, starting from Jacobian derivation for radar.

• Extended Kalman Filter
• Lidar and Radar Fusion
• High Accuracy

Depth Prediction Network

Modified Struct2Depth for indoor depth prediction, trained on a custom dataset for enhanced accuracy in environmental understanding.

• Depth Prediction
• Struct2Depth Modification
• Custom Dataset Training