PhD Student for Sustainable and Resource-Efficient Machine Learning at Linköping University

Machine learning has recently advanced through scaling model sizes, training budgets, and datasets; often at substantial computational and environmental costs. This PhD project targets sustainable and resource-efficient machine learning with a focus on methods that reduce compute, energy usage, memory and storage demands, and associated carbon emissions while aiming to maintain model quality.

The work will include developing new methodologies and algorithms for resource-efficient learning, for example via data selection and filtering (leveraging that not all data is equally informative). It will also investigate complementary approaches that reduce inference and deployment costs (e.g., model compression/simplification and hardware-aware optimization). We are also interested in how resource-efficiency interacts with broader sustainability aspects of machine learning such as robustness, fairness, and accessibility.

The position is formally based at the Division of Statistics and Machine Learning (STIMA) within the Department of Computer and Information Science at Linköping University in Sweden. The project will be carried out in a collaboration between STIMA (main supervisor: Assistant Professor Sebastian Mair) and the Sustainable Artificial Intelligence for Sciences (SAINTS) Lab (co-supervisor: Assistant Professor Raghavendra Selvan) at the Department of Computer Science of the University of Copenhagen in Denmark. We will strive for a tight collaboration between the groups.