Five PhD positions in Machine Learning @ University of Copenhagen

[Yevgeny Seldin]

The Department of Computer Science of the University of Copenhagen invites applications for five PhD positions in Machine Learning.

Application link: https://candidate.hr-manager.net/ApplicationInit.aspx?cid=1307&ProjectId=164839

Application deadline: 30 September 2025.

The five projects:

1. Green Computation Scheduling
   Computation has become the backbone of modern society, but it consumes a considerable amount of energy. The project aims to reduce carbon emissions from computation by designing algorithms to schedule computation at times of low carbon intensity of the electricity supply. The challenge is that due to high intermittence of green energy sources and multitude of independent consumers, prediction of carbon intensity is challenging. Even if it was possible to predict the supply, independent attempts to exploit low carbon energy would lead to demand spikes, invalidating the predictions. We aim to address this challenge by building on recent advances in online and reinforcement learning in adversarial environments, and further advancing this field of research. Further details about the project are available here. Candidates applying for this position are expected to have solid theoretical background and mathematical skills. Background in online learning, bandits, and theoretical reinforcement learning is an advantage. The project will be supervised by Yevgeny Seldin and Raghavendra Selvan. For inquiries concerning this project, please, contact Yevgeny Seldin <sel...@di.ku.dk>.
    
2. Resource efficiency for generative AI
   In this project, we will broadly investigate resource-efficient Large Language Models (LLMs) and their effect on the sustainability of AI. This can be at the level of developing novel algorithms, training, learning and prompting paradigms, or hardware optimization techniques that can result in reductions in the resources required when developing and deploying LLM pipelines.The interplay of resource efficiency with the broader sustainability of Generative AI (in terms of safety, fairness, and access) will be of particular interest. For more details about the project contact Christina Lioma <c.l...@di.ku.dk>.
   Candidates applying for this position must have strong skills in mathematics and ML, with a drive towards advancing the UN sustainable development goals. The project will be supervised by Christina Lioma, Maria Maistro and Raghavendra Selvan.  For inquiries concerning this project, please, contact Christina Lioma <c.l...@di.ku.dk>.
    
3. Sustainable Machine Learning for Earth Observation
   The PhD project will broadly investigate sustainable machine learning (ML) methods for Earth observation. This can include development of novel algorithms, learning paradigms, or hardware optimization techniques that can result in reductions in the resources required when developing and deploying ML pipelines for Earth observation. Specific areas of high interest include the development of high-resolution models from low-resolution labels and multi-modal data, as well as the creation of tailored quantization, data selection, and data condensation strategies specifically for Earth Observation data. The position is part of the TreeSense: Centre for Remote Sensing and Deep Learning of Global Tree Resources, and the candidate would closely work with collaborators at the Department of Geosciences and Natural Resource Management. Candidates applying for this position must have strong skills in mathematics and ML, with a drive towards advancing the UN sustainable development goals. The project will be supervised by Christian Igel, Ankit Kariryaa and Nico Lang. For inquiries concerning this project, please contact Christian Igel <ig...@di.ku.dk> or Ankit Kariryaa <a...@di.ku.dk>.
    
4. Resource-Efficient Machine Learning for Radiotherapy
   AI models will enable more precise and continuous adaptation of radiotherapy treatments to improve patients' prognosis. This project will develop algorithms to detect changes that occur during treatment and create individualised models that incorporate patients' medical history. We will develop new methods that can utilise the information that is available from existing treatment scans, but which currently cannot be fully utilised, The project’s goal is not only to improve treatment for the approximately 14,000 Danish cancer patients who receive radiation therapy each year, but also to develop solutions that can be scaled globally - particularly to benefit low- and middle-income countries. This is part of the AIM@CANCER project funded by the Novo Nordisk Foundation’s Grand AI Challenge. The project will be supervised by Jens Petersen and Raghavendra Selvan. For inquiries concerning this project, please contact Jens Petersen <ph...@di.ku.dk> or Raghavendra Selvan <rag...@di.ku.dk>.
    
5. Machine Learning for Design of Sustainable Food Processing
   Currently, plant ingredients are often refined to almost molecular purity - and then combined again to create structured foods. This isolation is resource intensive, and the removal of fibre and micronutrients can compromise the nutritional value. This can be mitigated by applying milder forms of processing that do not fully refine ingredients and leave some of the native structure of the plant material intact. These less refined ingredients however exhibit complex behaviour, and we therefore need machine learning to direct the experimental data generation that will be carried out by other team members in the project.
   Learning meaningful representation spaces that model the complex space of ingredients and their properties can be immensely useful. These representation spaces can be informed by multiple modalities of data, spanning time-series data, microscopy images, rheological measurements, and so on. Integrating these modalities into common representation spaces can help in the development of more sustainable food. The project will be in collaboration with Department of Food Sciences, UCPH. Further details about the project are available here.The project will be supervised by Christian Igel, Remko Boom and Raghavendra Selvan. For inquiries concerning this project, please contact Christian Igel <ig...@di.ku.dk>, Remko Boom or Raghavendra Selvan <rag...@di.ku.dk>.