Rakshith Sudhir Singh

About Rakshith Sudhir Singh

I\'m Rakshith \"Rak\" Singh, a Software Engineer and Computer Vision specialist currently…

Experience

1. Perception Systems Engineer ll

   Brunswick Corporation

   Jul 2021 — Present

   Core member of the Perception team in Brunswick\'s autonomy effort. Led the CVML Pipeline development - An intelligent and automated pipeline using Azure Cloud services for Data ingestion from a multimodal sensor suite, data storage, data processing, annotation and model deployment on vessels. So far, the pipeline has been executed and tested over several TBs of rosbag data and processed several thousands of images.\"Intelligence\" added to the pipeline using Object detection to filter extracted images for usefulness and GANs for up-scaling low-res images. Data transmitted from test vessels through LTE and Azure IoT Hub or collected as large rosbags (up-to several 100gb per rosbag)- Developed Brunswick\'s Marine Perception Ontologies and Object detection models for Marine autonomy use cases such as Captain Blindside monitoring, Obstacle avoidance etc. Strategized new techniques to augment data sets and managed data annotation and QA with labeling partners. Currently managing one of the largest closed-source Marine Perception data set- Managed over 10 students this past year, with varying experiences (PhDs to Freshmen) as part of my team to aide in R&D and develop potential talent pool for recruitment with a 80% team retention rate this year. Developed and maintaining cross-division collaboration across the org and university collaboration with Universities such as MIT (SeaGrant lab), UIUC etc- Developed Autonomy Cloud architecture for data capture from user vessels and OTA firmware and software updates. Working with external contractors for development of a small-footprint x86 based yocto-mender-ros-nvidia build with above mentioned capabilities- Leading the R&D effort of Brunswick\'s Marine Perception Stack - SLAM on water for large vessels (35+ ft), using various technologies and sensor configurations such as Radar based localization, Stereo Odometry, localization with GPS+RTK corrections, deep learning based VIO (Visual Inertial Odometry).