Ing. Varshini Muthukumar

• End-to-end development of customized AI solutions and deliver tailored, efficient solutions in logistic systems • Develop and improve perception algorithms for robotic perception • Design, train, validate,deploy and maintain ML models in production-on custom datasets for real time Object detection, Segmentation, Key point detection, OCR • Implemented pose estimation algorithms leveraging Apriltags technology to accurately determine the spatial orientation of objects in various applications.

Semantic Image Segmentation for Robotic Vision(Autonomous robot) • Developed a surface segmentation model for a delivery robot using cross-domain learning from Autonomous Driving datasets. • Conceptualised the perception stack for L3 level of automation for delivery robots to be operated on sidewalks • Researching and implementing state of the art publications in Semantic Image Segmentation, data-collection with the robot • Strategizing data-annotation tasks for the labelling team

Worked on improving 3D Point Cloud Denoising networks by including point pattern specifications • Explored SoTA deep neural networks (for e.g. PointCleanNet) to include point pattern specification in denoising of 3D objects defined by point clouds • Developed spectral characterization of Point patterns in surfaces defined by point clouds using Laplace-Beltrami Operator

Worked on Visual Object Tracking in videos using deep learning (SiamRPN++) • Developed a tool to create an unlabeled dataset of images of everyday objects which are of interest to a user(determined by the gaze data and video data of an eye tracker) • Explored various clustering techniques(Machine Learning) to create unlabeled image clusters which could be leveraged for training a network with Active learning • Experience with eye-tracking devices and real-world data

Worked on Automotive User Interfaces • Upskilled on basics of C# and Unity and created samples of User Interface designs for a Automotive Head-Up display • Testing of an application for driver monitoring and Assistance system

• Analyzed the feasibility of automated HMI validation and requirement specification of inhouse computer vision based tool • Experience in commissioning the Graphics Test Sytem(test automation) and being the SPOC for the team • System Testing of Instrument Cluster ECU for various Driver Assistance, Media and Traffic Sign Assist modules • Diagnostics and protocol level testing (CAN and BAP) • Experience in the V-Model/Agile software development and Automotive SPICE • ISTQB Foundation level certification

Compressed sensing system for efficient ECG signal compression at sub-Nyquist rates Sparse sensing of an ECG signal by sampling at sub-Nyquist rates, which results in high compression. The thesis explores the sampling matrices and reconstruction algorithms best suited for ECG signal.

Computer Graphics | Deep learning for Image Analysis | 3D Vision | Machine Perception

Garde 1.8 • Active participant in various technical workshops in campus. • Active volunteer in Outreach programmes conducted in village schools.