• Developed software modules for the excavation robot to selectively excavate regions of the rockface with high ore concentration.• Held responsibility as the interim lead for the software team of 12 until the new Head of Software was appointed.

• Implemented a perception module for the excavation robot arm to automatically scan the rock wall with a depth senor to generatea 3D map and compute the target orientation and position of the tool.• Developed a trajectory sampling module that generates a collision-free joint trajectory that satisfies task constraints.• Implemented a controller for the excavation robot arm to maintain tool orientation while applying force at the end-effector.• Led the integration of the robotic excavator system collaborating cross-functionally with mechatronics and UI teams.• Led the development of Autonomous Navigation software stack for a tracked mobile robot traversing very uneven terrains.• Developed ROS package for generating traversability cost map from LIDAR data. Dockerized this package along with ROS bridgenode to facilitate integration with the rest of the navigation stack implemented in ROS 2

• Implemented sampling-based Motion Planner for robotic arm manipulators equipped with excavation tools using OMPL library.• Created URDF and simulation models for diverse robots, encompassing arm manipulators with specialized tools. Enabled testingin simulation to catch software bugs before deploying on the real robot.• Developed a closed-loop controller for a tracked mobile robot to align itself against the rockface based on point cloud feedback.

• Developed an algorithm to identify optimal points of excavation by processing 3D point clouds of the rockface using PCL library.Demonstrated 2x improvement in the efficiency of the excavation subsystem.

• Collaborative Navigation: Developed a communication framework for sharing pose and visual landmark information among multiple Mobile units. Integrated this framework with the All-Source Positioning and Navigation pipeline (ASPN) and demonstrated 36% increase in navigation solution accuracy. • Cross-compiled the entire pipeline in Docker container and run on Nvidia TX2 boards.• Calibrated new IMU, Camera and Range radio and interfaced them with Nvidia TX2 board before conducting experiments on it.• Tuned the Inference based on a Probabilistic Graphical model of state variables and measurements from different sensors.

• Face Recognition with OpenCV on NVIDIA TK1 platform: Performed Data Augmentation and Image Registration. Trained and tested an SVM Classifier with different facial features (89% prediction accuracy). • Automatic Shot Boundary Detection: Developed and implemented an algorithm in OpenCV for detecting shot transitions in a video using Edge Tracking and DCT coefficients. Achieved less than 10% error rate.• Witness Video Summarization: Built a web application using Django, HTML5 and MySQL that captures video annotation data from Amazon Mechanical Turk workers. The application was hosted in Amazon Web services (AWS).

• Robotics: Designed and built a Tele-Presence robot.ROS packages for an Autonomous telepresence robot implementing SLAM and Motion planning algorithms and integrate them with Cisco’s DX-70 device. Developed an Android app integrating Cisco Video collaboration software with robot controls.• Analytics: Implemented an ARIMA model in Java backend for performing time series prediction on Supplier Business Ratings. Built a frontend tool using Angular JS and Chart.js to visualize the forecasts. • IoT: Built an Android App for Bluetooth Low Energy based identification system for conference rooms.

Intelligent control of semi – autonomous mobile vehicle:Objective: Construction and implementation of a Two-wheeled, Semi – Autonomous, BLDC motor driven mobile vehicle, that can be controlled from a remote location through a software application.Modelled and Simulated the system before implementing PID control for smooth transition in speed of BLDC motors.

3D Object Recognition and Pose Estimation:Developed a ROS (Robot Operating System) package to perform Object Recognition using Kinect sensor, Point Cloud Library, VFH global descriptors and FLANN for finding the closest match. The 6DOF pose of the object was estimated and sent to a Robotic- Arm controller. The entire system was implemented in (ROS).