Open calls in doctoral studies, the fourth period

Motivation letter

Please write a brief motivation letter (in English, maximum of 6000 characters with spaces) based on the following points:

1. Why are you interested in this PhD project, explain your choice.
2. What is your previous experience in this field? Explain how your educational and professional background relates to the project you are applying to.
3. What are the analytical/scientific methods you have practiced.
4. Describe briefly the methods and main results of your MSc thesis.
5. Decribe your earlier research activities, including research publications and conference presentations, if available.
Assessment criteria for motivation letter:

- motivation and argumentation of skills and the choice of the project
- relevant study and work experience and other relevant activities (publications, project management etc.) as required to present in the motivation letter.

Interview

The applicant must describe the wider scientific background of the doctoral project and possible applicability of the results, also their motivation to be admitted to PhD studies with particular project. The admissions interview is conducted by the admissions committee. Only applicants whose motivation letter is assessed positively will be invited to the interview (minimum positive result is 35 points out of 50).

The entrance interview is used to assess the following:

- knowledge of the wider scientific background of the project and possible application of the expected results
- applicant’s motivation to pursue doctoral studies in the relevant field of science and to work in this field
- wider analytical and generalization skills regarding the research and study topics.

International applicants who cannot be present at the interview in Tartu, may conduct an online interview. Applicants will be informed of their interview date and time by the respective faculty.

Both the motivation letter and entrance interview are assessed on a scale of 0 to 50 points, minimum positive score is at least 35 points. To be invited to an interview, the applicant must earn at least 35 points for the motivation letter.

Supervisors: Radwa Mohamed El Emam El Shawi, Stefania Tomasiello

AutoML tools are designed to automatically find the best machine learning models for specific tasks. However, they often lack the ability to incorporate domain-specific knowledge and provide explanations for their decisions. Integrating LLMs into AutoML can offer improved semantic understanding, enabling users to communicate their expertise and preferences in natural language. This fusion has the potential to create more powerful and user-friendly AutoML tools that bridge the gap between technical experts and machine learning novices, fostering better collaboration and knowledge exchange in the development of ML solutions. The goal of the doctoral work is to introduce approaches to combine the scalability and robustness of classical ML techniques with the vast domain knowledge embedded in large language models (LLMs). This allows for human-in-the-loop, interpretable AutoML.

Supervisors: Amnir Hadachi,Abdelaziz Bensrhair (INSA de Rouen),Paul Honeine (University of Rouen)

In the last decades, we witnessed rapid artificial intelligence advancements built upon deep learning (DL). Moreover, the DL decision mechanism is so obscure that testing is the only way to verify it. Hence, the process from training to testing any model is computationally demanding. Consequently, due to their high carbon footprint, DL networks become a concern for suitability. From this perspective, green learning (GL) has been presented as a potential solution to address these concerns. Thus, the Ph.D. topic is focused on exploring the possibilities of the GL paradigm and how it can be adopted in rethinking and redesigning the models’ architectures to reduce the carbon footprint of computer vision algorithms based on Deep learning.

Supervisors: Mart Loog, Ilona Faustova

Most cellular processes are regulated by protein phosphorylation. About 2% of eukaryotic genes code protein kinases, which phosphorylate at least 30% of the proteome. Although many kinasesubstrate pairs have been identified, in most cases it is not known how kinase specificity is achieved, to what extent and under what conditions a substrate is phosphorylated. The main goal of this doctoral thesis is to describe the specificity determinants of kinases Pho85, Mps1 and Cdc7. These are highly conserved kinases that control critical cell cycle processes. These kinases have been thoroughly studied by genetic methods, but we are lacking mechanistic understanding of their function. The Loog lab has extensive experience in cell cycle and phosphorylation studies, and this project plans to apply this to unravel the signalling mediated by these three kinases. Pho85p, Mps1 and Cdc7 are highly conserved, therefore the knowledge from yeast can be used to study these kinases in human cells. These kinases are central regulators of cell cycle processes and changes in their activity is often correlated with cancer. So far, kinase inhibitors for cancer treatment have been mainly designed to bind to the active site of the kinase. However, since kinase active sites are homologous and conserved, these inhibitors often have off-target effects. As kinase specificity seems to be mainly governed by distal docking interactions, the knowledge gained from these specificity studies can be used to design specific inhibitors to these central cell cycle kinases.