Personal profile

Research Focus

Artificial Intelligence (AI) based design and modelling of energy efficient computing systems

Computing systems have undergone a tremendous change in the last few decades with several inflexion points. Energy efficiency has now become a first order design parameter and constraint across the entire spectrum of computing devices. Many research surveys have gone into different aspects of energy efficiency techniques implemented in hardware and microarchitecture across devices, servers, HPC/cloud, data centre systems along with improved software, algorithms, frameworks, and modelling energy/thermals. This research project focuses firstly on the prediction of the behaviour of modern computing systems and the other angle is on the design of health monitoring systems that will be used as a case study. The safety of critical applications like autonomous driving and health monitoring which must operate under strict power budgets while maintaining reliable operation under dynamically varying operating conditions has become a major concern these days. Health monitoring systems must work reliably and under strict power budgets and various environments. This project will focus on the novel application of ML methods for predicting the behaviour of computing systems and predicting potential failures under any conditions. This project will then apply the developed models for the energy efficient operation of health monitoring systems. It will model the system’s power and reliability using ML/NNs and on implementing scalable ML/NN inference architectures based on the trained workload aware power/reliability models to help achieve low power while maintaining reliable operation.  The intension is to utilize supervised and/or unsupervised learning-based methods for building workload aware power and failure prediction models for processors. Data extracted from processors will be used for training and develop the machine learning models. New data protection schemes as well as adaptive power-reliability management policies will be developed and evaluated using as case studies emerging modules from autonomous driving and health monitoring applications where power and dependability are very critical design targets.