Vishnu Narayanan Moothedath

I am a doctoral candidate with a submitted thesis titled "Towards Efficient Distributed Intelligence: Cost-Aware Sampling and Offloading for Inference at the Edge". I work in the Information Science and Engineering (ISE) division under the School of Electrical Engineering and Computer Science (EECS) at KTH Royal Institute of Technology, under the supervision of Prof. James Gross and co-supervision of Prof. György Dán.

My current research focuses on enhancing the performance, energy efficiency, and responsiveness of future communication systems and edge AI using statistical guarantees and optimisation. One line of research aims to reduce the data captured, communicated and processed, achieved through (1) intelligent and cost-minimising sampling/sensing and (2) efficient inference offloading policies that selectively offload inference tasks in real time, based on various performance metrics including energy, latency, accuracy and the costs of offloading and asymmetric costs of misclassifications. A second line of research focuses on stochastic characterisation of 5G cellular networks in terms of delay violation probability, resource utilisation, and HARQ configurations with a specific focus on URLLC and industrial networks that need to meet service-level agreements.

Before starting my PhD in March 2021, I worked in the cellular industry for almost five years, collaborating with Intel and Apple in their respective LTE/5G-NR baseband modem groups, where I was responsible for several physical/MAC layer features, including Random Access. I completed my master's degree in Communication Systems from IIT Madras under the supervision of Professor Srikrishna Bhashyam and bachelor's degree in Electronics and Communication from NIT Calicut. As part of my Master's thesis, I worked on the optimum distributed beamforming on MISO downlink, where a novel algorithm was proposed to find the optimal transmit power allocation with a sum-power constraint.