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. 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.

Supervisor(s): Tuul Sepp, Raul Vicente Zafra, Mart Jüssi, Richard Meitern

The project focuses on applying artificial intelligence (AI) to improve the processing, analysis, and modelling of animal telemetry data in conservation and other applied contexts. Building on long-term and taxonomically diverse telemetry datasets — primarily more than 30 years of seal telemetry data, as well as studies on seabirds and garden dormice — the project aims to develop generalizable and transferable AI-based tools for analysing animal movement, habitat use, and behaviour. The project will use supervised machine learning for the biological annotation of telemetry data and deep learning to model complex non-linear relationships, enabling the automation of currently time-consuming data analysis workflows and the development of predictive models supporting conservation and development-related decision-making. The overall goal is to overcome existing bottlenecks in telemetry data analysis in spatial planning, species conservation, and environmental impact assessment, while developing decision-support tools for researchers, governmental agencies, and environmental practitioners. The project has a strong interdisciplinary dimension, integrating expertise from ecology, statistics, modelling, engineering, and computer science, while also fostering collaboration with international research and development partners.

Supervisor: Nadežda Kongi

The doctoral project will develop artificial-intelligence-assisted methods and scientific software for learning from electrochemistry data and using this knowledge in future electrochemistry, electrocatalysis, and materials research. The work will focus on newer experimental data from current institute research and accessible publication data, while also using earlier records where they help comparison and research memory.

A central result will be an AI-assisted electrochemistry research data and experiment-support platform. The platform will structure experimental records, preserve metadata and provenance, support data-quality and relationship analysis, retrieve relevant prior experiments, and help researchers plan new work using traceable evidence. The project will investigate how heterogeneous experimental data can be transformed into an AI-ready research asset, how systematic patterns and uncertainties can be identified, and how human-in-the-loop AI assistance can improve the reuse of scientific knowledge.

The research will be validated through focused case studies selected from the group's current and emerging research. Oxygen reduction reaction and carbon dioxide reduction provide strong starting contexts because the institute has accessible data and established research lines in these areas, while the approach may be applied to other suitable topics. The core emphasis is data systematization, data analysis, AI assistance, and reusable scientific software. The project is expected to produce scientific publications, reusable data and software outputs, and transferable methods for data-intensive experimental research.

Supervisor(s): Ivo Leito, Stefan Kuhn, Lauri Toom

The proposed PhD project aims to develop adaptive and computationally supported NMR methodologies for pharmaceutical structure elucidation. The work will combine advanced NMR experiments, complementary MS information, and AI-supported data analysis into integrated analytical workflows for impurities, degradation products, metabolites, and complex mixtures.

The central hypothesis is that adaptive workflows, in which intermediate interpretation guides subsequent measurements, can improve the efficiency and reliability of structure elucidation. The project will investigate multidimensional NMR methods such as HSQC, HMBC, TOCSY, DOSY, and non-uniform sampling in combination with MS and computational support tools.

Specific objectives include adaptive NMR acquisition workflows, integrated NMR-MS methodologies for complex mixture analysis, and experimentally guided closed-loop workflows combining analytical measurements and computational interpretation. Validation will be carried out using pharmaceutical case studies including impurity profiling and metabolite characterization. The project is expected to result in experimentally validated analytical workflows, reusable computational tools. The doctoral candidate will get interdisciplinary training in analytical chemistry, NMR spectroscopy, pharmaceutical analysis, and AI-supported data interpretation.

Supervisor(s): Heikki Junninen, Pilleriin Peets

This doctoral project utilizes AI-driven modeling alongside advanced soft-ionization mass spectrometry (IMS-VOCUS and AIM-VOCUS) to investigate atmospheric autooxidation pathways of volatile organic compounds (VOCs). The framework integrates artificial intelligence directly into data processing pipelines to automate multi-reagent peak identification, deconvolute overlapping isotopic envelopes, and resolve complex interface fragmentation patterns. By leveraging extensive datasets from the CERN CLOUD chamber, the project will construct predictive chemical models that map radical cascades under varying environmental conditions. A primary objective is utilizing this artificial intelligence framework to extrapolate laboratory observations to unexplored, low-pressure upper-tropospheric conditions. Ultimately, the project will deliver a scalable workflow that transforms complex, multidimensional mass spectrometry data into a mechanistic understanding of highly oxygenated organic molecule (HOM) formation.

Supervisor(s): Karl Kruusamäe, Robert Valner
Disaster response scenarios such as wildfires, earthquake aftermath, and search and rescue are among the most demanding operational environments. They are dangerous, time-critical, and inherently unpredictable. Deploying teams of autonomous robots (ground vehicles, aerial drones) alongside human operators can save lives by accelerating search coverage, providing real-time situational awareness, and reducing human exposure to hazardous conditions. However, orchestrating a heterogeneous robot fleet in the field remains a fundamental challenge. LLMs demonstrate strong capabilities in reasoning, natural language understanding, and step-by-step problem decomposition, making them compelling for translating a human commander's intent into structured multi-robot mission plans. This PhD develops AI-based methodologies for task planning and coordination in human-robot teams operating in unstructured environments, pursuing two objectives: (1) Advance LLM-based tactical planning for real-time multi-robot task allocation and coordination; (2) Develop strategic-level mission planning that grounds LLM reasoning in robot capabilities and world knowledge.

Supervisor(s): Kairit Sirts, Helen Uusberg
The proposed PhD project investigates how to develop more structured and
controllable AI-based psychological intervention systems, focusing particularly on
questions related to their architecture and evaluation. While recent advances in large
language models have largely solved the technical problem of generating coherent,
contextually appropriate, and empathic conversational responses, conversational
capability alone does not yet constitute an effective psychological intervention system.

The project focuses on three related research questions. First, it studies how
information about the user should be represented and maintained across repeated
interactions, and how such longitudinal user-state representations could support
intervention selection and adaptation over time. Second, it investigates how
psychological interventions can be implemented using modular architectures instead of
relying solely on monolithic prompts or fine-tuning approaches, as is done in many
existing systems. Third, it studies how adherence to intervention principles can be
automatically evaluated.

The project adopts a simplified experimental framework for studying these questions in
a laboratory or otherwise non-clinical setting. The studied interventions focus on
emotion regulation and include, for example, emotional validation, emotional
reappraisal, and cognitive restructuring. The project aims to develop a minimal
prototype system that enables studying these questions in a controlled setting. The
developed approaches are also potentially relevant for other conversational AI systems
with constrained goals, boundaries, or interaction principles.

Supervisor(s): Huber Flores, Zhigang Yin, Margit Kõiv-Vainik

Abstract: Food loss and waste are major contributors to climate change, hunger, and food insecurity. Addressing these challenges requires easy-to-use and scalable solutions that can monitor product quality and provide timely insights into problems across the food supply chain. This is particularly important for organic produce, such as fruits and vegetables, which have a short shelf-life and are prone to rapid quality degradation during production, transport, storage, and retail. Existing quality-estimation methods remain insufficient because they are often costly to deploy at scale, while manual visual inspection is still the dominant approach in practice. In this project, we investigate an AI-enabled solution that combines low-cost sensing with data-driven models for produce characterization and quality estimation. We envision to develop an approach that can support all stages of the supply chain while at the same time being inexpensive and easy-to-deploy.

Supervisor: Marina Lepp

This doctoral research investigates how different human–AI collaboration models influence workplace learning effectiveness in data engineering. Recent advances in generative AI have created new opportunities for supporting knowledge acquisition and problem-solving, yet current AI-assisted learning practices remain largely ad hoc and insufficiently understood. The project aims to develop and empirically validate a typology of human–AI collaboration models that differ in AI autonomy, human control, and distribution of cognitive responsibility. The research combines perspectives from generative AI, learning sciences, human–computer interaction, cognitive load theory, and organizational learning. Methodologically, the study combines a systematic literature review with an experimental evaluation involving professional data engineers learning unfamiliar technologies and solving practical engineering tasks using generative AI. Learning effectiveness will be assessed through knowledge transfer, task performance, cognitive load, and qualitative user experience measures. The research contributes new knowledge regarding how AI autonomy and human participation influence knowledge construction and sustainable skill development in technology-intensive work. The results are directly applicable to organizations seeking effective AI-supported workplace learning solutions and improved workforce adaptability in rapidly evolving technological environments.

Supervisor(s): Mark Fišel, Kalev Koppel

The aim of the project is to research approaches to automatic generation of sign language, with the goal of developing more accurate and reliable approaches to machine translation from text to Estonian Sign Language. Specifically, the project focuses on sign language generation

(1) processing pipeline and intermediate representation optimization and exploration,

(2) creation and use of synthetic data, and

(3) use of multilingual data and learning multiple sign languages in a single model, and the analysis and use of transfer learning.
The current work is part of an automatic sign language robot development project and will help increase accessibility for groups of people whose communication is hindered.

The candidate is expected to have

- good writing and communication skills in Estonian and English,

- a Master's degree in computer science/informatics, data science, computational linguistics, or a related field

- very good knowledge of Python and the transformers library, as well as sufficient knowledge of modern machine learning to process text, audio, and video data, train and test models

- curiosity and readiness to engage in exploratory research

Supervisor(s): Jaan Aru

How can we develop AI systems that support learning? The Estonian state has launched the AI Leap program, but its impact is currently unclear. This doctoral thesis investigates the impact of AI Leap, where one component is also an AI-based learning application. Within this project, a system is being developed that enables, through automated analysis, the study of the impact of an AI-based tutor on a student’s behavior. This system will be validated and applied using data from AI Leap, where these data will be analyzed to understand which aspects of the learning application’s behavior supports the learning process. Over the course of the project, a better understanding of AI Leap’s impact will emerge. The project will also produce recommendations on how to use AI-based tutors more effectively. The project has broad international relevance and visibility.

Supervisor(s): Kadri Leetmaa, Bradley Loewen, Kadri Kangro

The proposed doctoral research examines the role of enterprise networks in enabling and sustaining social innovation in peripheral rural regions. While rural areas are increasingly recognised as crucial for addressing major societal challenges—including climate adaptation, biodiversity restoration, food security, renewable energy transitions and territorial resilience—they simultaneously face demographic decline, service withdrawal and weakening institutional capacity. Existing research has primarily focused on community-led initiatives, public sector interventions and entrepreneurship, while the contribution of enterprises and business networks to rural social innovation remains insufficiently understood. This project addresses this gap by investigating how enterprises contribute to social innovation through collaboration, knowledge exchange and the provision of essential services, thereby strengthening the resilience of peripheral rural communities. The research advances theoretical understanding at the intersection of rural studies, social innovation and economic geography by conceptualising enterprise networks as key actors in rural innovation ecosystems. Empirical research will be conducted within the Horizon Europe REROOTED project, applying a common international methodology to compare frontrunner rural communities across participating European countries.

Supervisor(s): Anneli Kährik, Anto Aasa, Tiit Tammaru

The doctoral project focuses on developing AI-based spatial forecasting methodologies and Explainable Artificial Intelligence (XAI) approaches for analysing socio-spatial urban change and the impacts of planning decisions in urban regions. Contemporary urban planning still relies largely on fragmented datasets and static impact assessments that are unable to capture the cumulative and long-term effects of housing development, mobility transitions, and planning decisions on housing affordability, segregation, displacement risks, and access to services.

The project integrates spatial, population, mobility, housing, and planning datasets into an AI-based socio-spatial forecasting framework capable of modelling future urban development trajectories and estimating the impacts of planning scenarios before implementation. Particular attention is given to how housing developments, urban regeneration, and mobility transformations influence socio-spatial inequalities and accessibility in urban regions.

The dissertation is both interdisciplinary and cross-sectoral. It combines approaches from human geography, urban studies, spatial data science, mobility studies, and AI-based modelling, while also integrating practical public- and private-sector needs and real planning cases into the modelling framework. Methodologically, the dissertation applies spatial regression methods, spatial machine learning approaches (e.g. XGBoost, LightGBM), mobility analytics, and Explainable AI techniques, including SHAP-based model interpretation.

A key contribution of the project lies in applying Explainable AI to urban and regional research in order to make AI-based forecasting models more transparent, interpretable, and usable in planning and public-sector decision-making. The dissertation contributes to urban geography, spatial data science, and AI-supported planning research by developing interpretable AI methodologies for modelling socio-spatial urban change and supporting more evidence-based and socially sustainable urban planning practices.

Supervisor(s): Valdis Laan

This project is in the field of algebra, more precisely in semigroup theory. The aim of the project is to study the properties of certain naturally defined categories of acts over semigroups and compare those results with corresponding results about categories of modules over rings without identity. We would like to understand if these categories could be more suitable for defining Morita equivalence relation on the class of all semigroups compared to the category of all firm acts. The categories that we wish to examine contain all closed acts or all unitary and nonsingular acts. The problems that we will study include the following: are these categories reflective subcategories in the category of all acts, how are the limits and colimits computed in these categories, what are the projective and free objects, do they have a (co)generating set of objects and so on.

Supervisor(s): Hannes Kollist, Pawel Roszak, Yuh-Shuh Wang

Plant cell walls are dynamic structures that continuously adjust their composition in response to developmental and environmental cues. Callose is a β-1,3-glucan polysaccharide that plays key roles in cell plate formation, pollen tube growth, and plasmodesmata regulation during phloem differentiation. Beyond its developmental functions, recent evidence indicates that ectopic incorporation of callose into tree secondary cell walls can reduce biomass recalcitrance and improve enzymatic accessibility, offering new opportunities for wood processing. Despite its importance, the regulatory mechanisms controlling callose deposition, localization, and turnover remain poorly understood.

We have identified a set of transcription factors that promote callose formation. When expressed in their native context, these factors regulate callose deposition in maturing phloem sieve elements; when ectopically activated, they induce a broader genetic program sufficient to trigger callose accumulation at cell walls and plasmodesmata. Transcriptomic analyses of loss- and gain-of-function lines have generated a comprehensive list of candidate regulators potentially controlling callose synthesis, targeting, and degradation.

This PhD project aims to dissect the molecular network governing callose dynamics. We will (i) functionally characterize key candidate genes required for callose deposition in developing sieve pores and plasmodesmata; (ii) perform a genetic screen for the suppressors of NAC020 overexpression phenotype (ectopic callose deposition) and characterization of the identified mutants; (iii) integrate promising regulators into a VND7-driven xylem transdifferentiation system to evaluate their capacity to modulate callose accumulation in secondary cell walls; and (iv) establish a scalable Arabidopsis xylem cell culture platform to test natural genetic variants identified from Nordic-Baltic birch populations for their impact on cell wall composition and wood biomechanics.

Together, this work will elucidate the mechanisms controlling callose deposition and leverage this knowledge to engineer improved wood properties and biomass quality for sustainable bio-based applications.

Supervisor(s): María José Guzmán Monsalve, Laxmipriya Pati (University of Salamanca, Spain)

The advent of gravitational wave astronomy after the first detection of a binary black hole merger has imparted a significant momentum to the field of numerical relativity and the numerical modelling of gravitational theories beyond general relativity. The theoretical framework to model modified numerical relativity is almost restricted to metric-based theories of gravity, and its understanding for theories based on the tetrad or the linear connection is far from complete. Therefore, it is of interest to develop the theoretical foundations of numerical relativity in a formalism where the tetrad field is used instead of the metric. The use of the tetrad as the fundamental field can offer various advantages, ranging from its better suitability for defining observable physical quantities, to its potential to simplify the split of time and space from the equations of motion. A crucial aspect to be inquired into corresponds to the potential of the tetrad formulation to ease the assessment of strong hyperbolicity of the system of differential equations dictating the behavior of spacetime, a crucial requirement for stable numerical simulations. It is intended to develop this branch of research for tetrad-based general relativity and the teleparallel equivalent of general relativity.

Supervisor(s): Mihkel Pajusalu, Aditya Savio Paul

The aim of this doctoral thesis project is to develop and test a self-supervised deep learning framework to help planetary rovers and military UGVs navigate in unstructured environments (i.e. environments that are not regular city environments or roads) using passive and proprioceptive sensor systems.
Geometry and occupancy mapping are learned from camera data, while traversability properties are learned from proprioceptive signals such as IMU and wheel encoders. The resulting machine learning model would construct a discrete geometrical 3D representation of the surroundings with embedded traversability information.
The framework will be evaluated in simulated and real-world off-road environments representative of possible extra-terrestrial and off-road applications.