Sai Teja Meda

Flashed base software and ran OTA tests on infotainment benches Tested WLAN connectivity, media streaming, and cloud features post-update Conducted Bluetooth IOP (MAP, HFP, PBAP, A2DP) and Android Auto/CarPlay tests Performed media/network testing for calls, ensuring in-vehicle integration Supported test plans and JIRA traceability Created AI automation scripts with Appium/Selenium, boosting efficiency 50% Analyzed diagnostics and protocols with CANoe and UDS for ECU reliability

Robust 3D depth & motion estimation using unsupervised learning on video datasets. Developed unsupervised model for monocular depth & egomotion estimation in visual odometry. Integrated intrinsic camera parameter prediction for robust calibration of uncalibrated cameras. Used PyTorch Lightning on GPU clusters (CUDA) to train on KITTI & Ego4D datasets. Optimized ResNet encoder-decoder with multi-scale losses and auto-masking for improved depth and motion accuracy.

• Developed risk-controlled Prediction Sets (RCPS) technique to ensure ≤5% miscoverage on Frame prediction dataset • Implemented RCPS for multi-class classification, achieving 95% coverage on unseen test data • Designed and implemented a proof of concept (PoC) for the RCPS method from scratch