Nagarjun PR

Master’s student in Artificial Intelligence at Brandenburg University of Technology Cottbus-Senftenberg, focusing on efficient and reliable deep learning systems.My Master’s thesis focuses on adaptive early-exit mechanisms for deep neural networks, especially transformer-based models. I investigate how dynamic exit decisions, threshold adaptation, and reliability-aware inference can reduce computational cost while maintaining accuracy and robustness. The goal is to develop an efficient framework for making neural network inference faster and more resource-aware.