Jeethesh Pai Umesh

• Worked with urban scene understanding team on the robustness of auto labeling • Topic: “ Automatic label error correction for 2D object detection • Researched several state-of-the-art Transformers and other localization approaches. • Experimented with several backbones and our approach using Py T orch, Python, Microsoft Azure GPU & Visual Studio Code Tools Microsoft Azure, PyTorch, OpenCV, Python, Linux

• Involved in the development of interior vehicle sensing algorithm • Built a robust semantic segmentation network with mIoU of 79% on the real dataset • Converted the model to ONNX format and deployed on Ambarella SoC using ROS2 & C++ • Familiarized with ROS2 and data acquisition with the camera on Edge device • Organized 70% of codes in an Agile environment using JIRA & Bitbucket , PyCharm

• Developed two computer vision related python applications for autonomous driving and deployed using Docker on the Linux • Acquired data from sensors installed on research locomotive using ROS, Python, C++ • Supported in CI & CD with test-driven development in Git repository using PyCharm and Visual Studio and effectively communicated between teams • Achieved an accuracy of 92% on weather classification using ResNet 50 and improved training by resolving the data imbalance using TensorFlow and Python.