Skip to content

Admissions

We offer the possibility to join world-class research groups either at Aalto University or University of Helsinki, with multiple interesting research projects to choose from. We organise two open calls for new doctoral students annually, typically in the winter and summer.

Our fully-funded doctoral student positions have gained a lot of international interest since their introduction in 2014. We appreciate applicants with diverse backgrounds and enthusiasm in computer science.

The quality of research and education in both HICT universities is excellent and our doctoral students are typically hired as full-time employees for the duration of their doctoral studies. The average duration of doctoral studies is four years in Finland.

You are most welcome to apply for HICT doctoral student positions! Our open positions are announced on this webpage.

Summer call 2022 – How to apply?

  1. Get familiar with the application process, eligibility requirements and compulsory attachments described on General HICT Call FAQ.
  2. Choose the projects from the list below.
  3. Fill in the application form and send it before deadline.

Our Summer Call is now open. The application form closes on 28 August at 11:59 pm Finnish Time (UTC+3)


Helsinki ICT network: Doctoral student positions in computer science

The Helsinki Doctoral Education Network in Information and Communications Technology (HICT) is a joint initiative by Aalto University and the University of Helsinki, the two leading universities within this area in Finland. The network involves at present over 80 professors and over 200 doctoral students, and the participating units graduate altogether more than 40 new doctors each year.

The participating units of HICT have currently funding available for qualified doctoral students. We offer an exciting opportunity to join world-class research groups, with over 20 research projects to choose from. The activities of HICT and the themes of open positions are structured along five research area specific tracks:

  • Algorithms and machine learning
  • Life science informatics
  • Networks, networked systems and services
  • Software and service engineering and systems
  • User centered and creative technologies
     

If you wish to be considered as a new doctoral student in HICT you can apply to one or a number of doctoral student positions. We actively work to ensure our community’s diversity and inclusiveness. This is why we warmly encourage qualified candidates from all backgrounds to join our community.

The online application form closes August 28, 2022  at 11:59pm Finnish time (23:59 EET, Eastern European Time UTC+3).

Our researchers are seeking doctoral candidates to fulfill positions for 21 projects. They are listed below grouped by primary research area.

How to Apply: Read through the list of available positions and select the topic(s) that you are interested in. Click on the link of the topic that you are interested in, select Apply and Create account. (Note: you will only have to create an account for your first application).

Projects in the HICT Summer call 2022

Algorithms and Machine Learning Research Area Projects

Life Science Informatics Research Projects

Networks, Networked Systems, and Services Research Projects

Choose your topics below and fill in their application form.

Algorithms and Machine Learning Research Area Projects

AI-Powered Simulation, Optimization and Inference

Supervision: Profs. Luigi Acerbi (University of Helsinki); Jukka Corander (University of Helsinki), Arno Solin (Aalto University), other professors involved in the topic.

Research Area: Algorithms and Machine Learning (AML)

Recent advances in machine learning have shown how powerful emulators and surrogate models can be trained to drastically reduce the costs of simulation, optimization and Bayesian inference, with many trailblazing applications in the sciences. In this project, the candidate will join an active area of research within several FCAI groups to develop new methods for simulation, optimization and inference that combine state-of-the-art deep learning and surrogate-based kernel approaches – including for example deep sets and transformers; normalizing flows; Gaussian and neural processes – with the goal of achieving maximal sample-efficiency (in terms of number of required model evaluations or simulations) and wall-clock speed at runtime (via amortization). The candidate will apply these methods to challenging problems involving statistical and simulator-based models that push the current state-of-the-art, be it for number of parameters (high-dimensional amortized inference), number of available model evaluations (extreme sample-efficiency) or amount of data. The ideal candidate has expertise in both deep learning and probabilistic methods (e.g., Gaussian processes, Bayesian optimization, normalizing flows).

References:

Keywords: Machine learning, emulators, amortized inference, Bayesian optimization, normalizing flows, simulator-based inference

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Bayesian Workflows for Iterative Model Building and Networks of Models

Supervisor: Aki Vehtari (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

Statistical analysis is critical when it comes to obtaining insights from data. Despite the practical success of iterative Bayesian statistical model building, it has been criticized to violate pure Bayesian theory and that we may end up with a different model had the data come out differently. In this project, we formalize and develop theory and diagnostics for iterative Bayesian model building. The practical workflow recommendations and diagnostics guide the modeller through the appropriate steps to ensure safe iterative model building, or indicate when the modeller is likely to be in the danger zone.

Sometimes the user wants also to build and compare models of different complexity or based on different assumptions.  Similar workflow ideas can be used to support also analysis of networks of models, making it easier to illustrate similarities and important differences. By making applied scientific research and data analysis more reliable and reproducible, our understanding of the world and decision-making will be improved. Related paper

https://arxiv.org/abs/2011.01808 and video https://www.youtube.com/watch?v=ppKpwtGy8KQ


Deep generative modeling for continuous-time dynamics and biomedicine

Supervisor: Assoc. Prof. Harri Lähdesmäki (Aalto University)

Research Areas: Algorithms and Machine Learning (AML), Life Science Informatics (LSI)

We are looking for a doctoral student to work on deep generative modeling for applications that are at the intersection of biomedicine and health as well as continuous-time dynamics. Our research work involves developing new deep generative modeling methods for large-scale health datasets from Finnish biobanks, computational immunology, various single-cell datasets, and general data-driven modeling methods for continuous-time dynamics. Methodologically, the work revolves around e.g. conditional deep generative models, neural ODEs, and Bayesian methods. Work can focus more on method development or can also include applications, depending on applicant’s own preference. Applicants are expected to have good knowledge of machine/deep learning, statistics, programming, and (optionally) interest in bioinformatics and biomedicine. For more information, see our research group web page: https://research.cs.aalto.fi/csb/


Energy-efficient high-performance computing ecosystems

Supervision: Maarit Käpylä and Linh Truong (Aalto University)

Research Areas: Algorithms and Machine Learning (AML) and Networks, Networked Systems and Services (NNSS)

The energy consumption of large-scale computations and data analysis tasks is huge, but the current mindset of users and software developers for applications run on modern high-performance computing (HPC) platforms is not directed towards re-usability nor sustainability aspects. We are developing an ecosystem of software tools and services to enable HPC applications to use current and future heterogeneous HPC platforms in an optimal, energy-efficient and sustainable manner  through optimizing the HPC solutions, developing intelligent methods to avoid the data  bottleneck, harnessing the full potential of cloud computing in the analysis of huge data sets, and by integrating this into a seamless HPC ecosystem that performs the desired tasks in an automated, fault-tolerant, and reusable way. All these aspects are critical in enabling scientific applications at the Exa-scale, where Europe is heading currently with the in-auguration of the two pre-exascale supercomputers, one of which is LUMI, our target platform at CSC, Finland. We are looking for a PhD candidate with education and preferentially some experience in big data analysis platforms and their distributed machine learning solutions and/or heterogeneous distributed HPC systems.The PhD project will require contributions to both domains. 


Explainable AI for virtual laboratories

Supervision: Profs. Kai Puolamäki (University of Helsinki), Arto Klami (University of Helsinki)

Research Area: Algorithms and Machine Learning

FCAI is actively developing methods and software for virtual laboratories to enable AI assistance of the research process itself. Efficient human-AI collaboration requires methods that are either inherently capable in providing explanations for the decisions, or methods that can explain decisions of other AI models. For instance, the user needs to know why AI is recommending a particular experiment to be conducted or why AI is predicting a particular outcome, and they should always be aware of the reliability of the AI models. We are looking for a candidate that can conduct research on explainable AI and uncertainty quantification. The project will be conducted in a team consisting of AI researchers with access to researchers specialized in various application areas. The applicant should be interested in incorporating the techniques as part of general virtual laboratory software developed at FCAI for broad applicability. 

Keywords: Virtual laboratory, explainable AI, uncertainty quantification, human-AI collaboration

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Explainable and robust AI for scientists

Supervisor: Kai Puolamäki (University of Helsinki)

Research Area: Algorithms and Machine Learning (AML)

Machine learning and AI are extensively used in the sciences. When modelling physical systems, understandability and statistical robustness of the models is often more important than predictive accuracy. We are looking for talented postdoctoral researchers and doctoral students to study explainable and understandable AI and to study the uncertainty quantification of the AI models. While the AI methods we develop are generic and not tied to any specific application domain we work closely with scientists to build Virtual Laboratory for Molecular Level Atmospheric Transformations (https://wiki.helsinki.fi/display/VILMA). More information: https://bit.ly/edahelsinki2022jobs 

Keywords: explainable AI, uncertainty quantification


Intrinsic motivation-driven user modeling

Supervision: Prof Christian Guckelsberger (Aalto, primary supervisor); Prof Perttu Hämäläinen (Aalto, co-supervisor)

Research Area: Algorithms and Machine Learning (AML)

AI-assisted decision making requires human-centric AI capable of inferring a user’s motivations and predicting accurately how their experience and behavior will change as outcome of a decision on either side. There is common agreement amongst cognitive scientists that much of our behavior and experience is not only driven by separate consequences or instrumental outcomes, but also by intrinsic motivations [1]. Crucially though, despite offering important benefits such as domain independence, computational models of intrinsic motivations have not been extensively leveraged for user modeling. This project will push this agenda further by addressing amongst others, what constitutes psychologically plausible models of intrinsic motivation, investigating which model can serve as a predictor for certain types of experience and behavior, and inferring the best model from interaction with the user. The project sets out to tackle these challenges for player modeling in videogames as quintessential intrinsically motivating activities. It will then translate the insights into other domains of human-computer interaction. The supervisors have established the basis for this work through pioneering qualitative and quantitative proof-of-concepts [2,3] as well as theoretical studies [4].

The postdoc or PhD student will design, implement and execute studies to push the state-of-the-art of user experience and behavior modeling. A strong candidate will have solid coding experience, good knowledge of deep reinforcement learning and an interest in cognitive modeling and videogames. Prior experience in conducting user studies is an asset.

References:
[1] Ryan & Deci. (2000). Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being. American Psychologist, 55(1), 68–78.
[2] Guckelsberger, Salge, Gow & Cairns. (2017). Predicting Player Experience Without the Player. An Exploratory Study. Proc. CHI Play, 305–315.
[3] Roohi, Guckelsberger, Relas, Heiskanen, Takatalo & Hämäläinen. (2021). Predicting Game Difficulty and Engagement Using AI Players. Proc. CHI Play, pp.1-17.
[4] Roohi, Takatalo, Guckelsberger & Hämäläinen. (2018). Review of Intrinsic Motivation in Simulation-Based Game Testing. Proc. CHI, 1–13.

Keywords: intrinsic motivation, user modeling, reinforcement learning, human-computer interaction, human-centric AI, cognitive science, videogames

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Learning to handle rare events in autonomous driving

Supervision: Juho Kannala (Aalto University); Alexander Ilin (Aalto University), Joni Pajarinen (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

In autonomous driving, rare events are events that are typically not encountered during simulation and very rarely in the real world. What makes rare events challenging is that they cannot be exactly simulated due to the huge variety of different possible events and the lack of real world data. However, as evidenced by human drivers, responding reasonably to rare events is possible. To cope with rare events this project focuses on learning conditional object-centric representations from unlabeled computer vision data and using these representations in models that can be quickly updated and conditioned on the current driving context. Further, the models will be extended into an adaptive contextual replanning framework that allows for fast response with non-stationary models. The learned models and replanning framework will be made robust in a simulation environment where adversarial agents will learn to cause rare events. An ideal candidate has background knowledge in computer vision, deep learning, or/and reinforcement learning. Depending on the background, a successful candidate may focus on all of the proposal parts above, or, only a subset.

Keywords: Autonomous driving, computer vision, deep learning, reinforcement learning, planning

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Learning with probabilistic principles

Supervisor: Arno Solin (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

We are looking for exceptional and highly motivated doctoral students to work on applications and methods development in combining probabilistic methods with deep learning, real-time inference, and dynamical models. The project relates to approximate inference and learning in systems with non-linear dynamical priors. Possible application areas are in sensor fusion and computer vision.

The work will be done in close collaboration with the supervisor and other members of the team at Aalto University. Doctoral students in the group are encouraged to make research/internship visits to collaborating universities/companies. Successful candidates are expected to have completed a Masters’s degree and have familiarity with machine learning and statistics.

For more information on the broader research scope and recent publications and pre-prints, see the supervisor’s home page at http://arno.solin.fi

 


Maximally Autonomous AI Assistant (MAMAA)

Supervision: Samuel Kaski (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

Current machine learning approaches are showing outstanding success in automating tasks. However, they require the goal to be specified precisely, for instance as rewards, and in real-world applications we often cannot do that. Then reinforcement learning based solutions will happily produce complex solutions which we do not want, for instance in robotics or drug development pipelines.

In this project, we develop inference methods for AI assistants which automate as much as possible but not more – they need to get more information from the human when necessary, to reduce their uncertainty about the goal and to acquire generalisable long-term decision-making skills through reinforcement learning to reach this goal.

We are looking for a PhD student knowledgable in probabilistic machine learning and reinforcement learning. Additional knowledge in any of the following will be helpful: game theory, multiagent systems,  computational rationality, inverse reinforcement learning.

This project is collaboration with Prof. Ville Kyrki (robotics), TU Delft, MIT, pharma and self-driving car companies


Machine learning for drug design

Supervision: Samuel Kaski (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

Recent progress in machine learning for generative and predictive models of molecules has made computational drug design possible and created digital twins for the drug design Design-Make-Test-Analyze (DMTA) cycle, leaving as bottleneck the human skills needed for the Analyze step for deciding which molecule to make next. We  develop methods and models for cooperative AI assistants for the Analyze step to be able to help designers operate a virtual drug design laboratory. The principles of virtual laboratories, and cooperative AI-assistance in them, will be generalizable to other important research fields such as material science.

We are looking for motivated candidates with background in computational sciences, machine learning, statistics. This is a joint research project between Aalto, Chalmers University and AstraZeneca. You will join the  Probabilistic Machine Learning (PML) research group at Aalto, https://research.cs.aalto.fi//pml/

Supervision: Professor Samuel Kaski (Aalto), Dr Ola Engkvist (Chalmers University), Dr Markus Heinonen (Aalto)

Keywords: probabilistic modelling, human-in-the-loop modelling, drug design, deep learning


Multi-level simulation for sustainable autonomy

Supervision: Profs. Laura Ruotsalainen (University of Helsinki), Ville Kyrki (Aalto), Joni Pajarinen (Aalto)

Research Area: Algorithms and Machine Learning (AML)

To study future sustainable mobility, FCAI has built Sustainable Autonomous Mobility Virtual Laboratory. The virtual laboratory will allow studying effects of autonomous traffic starting from control of individual vehicles, to their environmental effects such as pollution and noise, as well as their socio-economic effects. The virtual laboratory will integrate several simulators including an autonomous vehicle simulator, as well as other simulators modeling relevant phenomena. A central challenge in the integration is the exchange of information between the individual simulation models with different parameterizations and objectives. We approach this as an AI challenge where parameters of all simulators are inferred jointly from pools of data for each model. The somewhat conflicting objectives of different simulations require the development of multi-objective multi-agent reinforcement learning simulations.

Keywords: Multi-level simulation, sustainability, autonomous vehicles, multi-objective reinforcement learning, multi-agent reinforcement learning

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Multivariate variable selection with applications to genetics

Supervisor: Matti Pirinen (University of Helsinki)

Research Area(s): Algorithms and Machine Learning (AML) ; Life Science Informatics (LSI)

Our current ability to measure massive numbers of variables, for example, in biological sciences (millions of DNA variants) or in personal health applications (thousands of time points in activity monitoring) call for efficient ways to separate the important variables from unimportant ones. Additionally, the outcome space of the variable selection problem may be multidimensional, e.g., we may search for DNA variants that affect both measured biomarker levels and disease risk. This setting calls for methods to decompose the multivariate outcome space into most informative components.

We are looking for a PhD student to work on realistic multivariate variable selection problems where variables may have been measured incompletely and where we may have access only to statistical summaries of the underlying data. Our previous work include the Bayesian approach implemented in the FINEMAP algorithm (www.finemap.me). We have also first-hand experience on large-scale genetic data analyses, including, e.g., the world’s largest study on migraine genetics (Nature Genetics 2022, 54: 152-160).

This project includes both theoretical (statistical) and empirical (simulations and genetic data analysis) components and is connected to the most recent resources of genetic and health-related data from Finland and across the world.

Research group pages: https://www2.helsinki.fi/en/researchgroups/statistical-and-population-genetics


Next-generation likelihood-free inference in ELFI

Supervision: Jukka Corander (University of Helsinki); Luigi Acerbi (University of Helsinki)

Research Area: Algorithms and Machine Learning (AML)

ELFI (elfi.ai) is a leading software platform for likelihood-free inference of interpretable simulator-based models. The inference engine is built in a modular fashion and contains popular likelihood-free inference paradigms, such as ABC and synthetic likelihood, but also more recent approaches based on classifiers and GP emulation for accelerated inference. We are looking for doctoral students, postdoctoral researchers and research fellows to spearhead development of the next-generation version of the inference engine supporting new inference methods, including the use of PyTorch and deep neural networks for amortized inference, and using ELFI in cutting-edge applications from multiple fields of science. The ideal background requires programming experience with modern deep learning frameworks (e.g., PyTorch) and familiarity with probabilistic inference and simulator-based inference.

Keywords: Machine learning, emulators, likelihood-free inference, simulator-based inference

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Privacy in machine learning for health

Supervisor: Antti Honkela (University of Helsinki)

Research Area(s): Algorithms and Machine Learning (AML)

Differential privacy allows developing machine learning algorithms with strong privacy guarantees. In this project, we will study the privacy of state-of-the-art deep learning models used in health applications. We will seek to both evaluate the level of privacy provided by standard approaches, as well as apply differential privacy to improve the privacy. The project requires a background in machine learning and ideally deep learning. Background on privacy is an advantage but not required.

More information and papers: https://www.cs.helsinki.fi/u/ahonkela/ or by email (antti.honkela@helsinki.fi).


Privacy-preserving data sharing and federated learning

Supervision: Antti Honkela (University of Helsinki), Samuel Kaski (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

Many applications of machine learning suffer from limited training data availability because data holders cannot share their data. The aim of this project is to develop solutions to this fundamental problem through privacy-preserving data sharing using differentially private synthetic data as well as through efficient privacy-preserving federated learning methods. The security and privacy will be guaranteed by a combination of differential privacy and secure multi-party computation.

In this project, you will join our group in developing new learning methods operating under these guarantees, and applying them to real-world problems. Collaboration opportunities will enable testing the methods on academic and industrial applications. A strong candidate will have a background in machine learning or a related field. Experience in privacy-preserving techniques such as differential privacy or secure multi-party computation is an asset.

Keywords: differential privacy, federated learning, synthetic data

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Probabilistic multi-agent modeling for collaborative AI assistants

Supervision: Samuel Kaski, Frans Oliehoek (TU Delft), other FCAI professors

Research area: Algorithms and Machine Learning (AML)

We study how to build collaborative assistants which are able to help another agent perform their task. The assistant does not know the agent’s goal in the beginning and has to learn it as a part of this “zero-shot” assistance scenario.

This is interesting as a fundamental multi-agent modeling problem, and in building collaborative AI assistants for human-AI research teams in decision making and design, formulated as sequential decision making. We are looking for a researcher interested in developing with us the theory and inference methods for this new task, or applying the assistants with other FCAI researchers to solving tough decision making and design tasks.

The work will involve probabilistic modeling, multi-agent formulations, POMDPs and reinforcement learning, and inverse reinforcement learning.

Keywords: probabilistic modeling, multi-agent formulations, POMDP, reinforcement learning, inverse reinforcement learning

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Reconstructing Crisis Narratives research project

Supervisor: Nitin Sawhney (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

Join a collaborative team of PhDs, Postdocs and academic researchers working on a research project, Reconstructing Crisis Narratives for Trustworthy Communication and Cooperative Agency, supported by the Academy of Finland. The project, jointly conducted between Aalto University and the Finnish Institute for Health and Welfare (THL), is seeking to analyze and reconstruct crisis narratives using mixed-methods, combining qualitative research for narrative inquiry with computational data analytics of crisis discourses in news and social media to understand global pandemics. Candidates will work at the intersection of Human-Computer Interaction (HCI), design research, computational social sciences, and public health for critical societal impact. We expect the candidates to have backgrounds in computer science and/or the social sciences.

Potential duties and tasks may include the following:
– Automating content analysis for narrative work using suitable machine learning techniques for Natural Language Processing (NLP) such as Conversation Analysis, Content Classification, and/or Sentiment Analysis. This includes collecting and curating datasets, devising suitable methodologies, setting up the research infrastructure and tools, and a pipeline for data extraction, analysis and validation.

– Representing and visualizing crisis narratives to support understanding and collaborative sensemaking among key stakeholders and diverse publics. These tools should support browsing, searching, and exploring information for selected crisis themes and narratives. Work in this area includes not only developing prototypes of visualizations, but also conducting design research, user experience (UX) evaluation, and pilot assessment of such tools.The candidate is not expected to master all these domains, but work closely with a multi-disciplinary research team to lead design and development efforts, while learning and contributing to ongoing work in specific research areas of interest. 


Representation learning for geometric computer vision

Supervision: Juho Kannala (Aalto University) and Arno Solin (Aalto University)

Research Area: Algorithms and Machine Learning (AML)

Machine learning based approaches have enabled progress in many classical problems of geometric computer vision, such as stereo depth estimation, image-based 3D modeling and visual localization. One example of learnt models are deep convolutional neural networks, which can provide useful priors for under-determined problems such as stereo depth estimation. Another recent example is learning of neural radiance fields (NeRFs), which are implicit scene representations with potential applications in problems such as new view synthesis, image-based modeling and visual localization. In this project, the aim is to focus on developing learning based approaches for geometric vision problems, which are relevant for machine perception and autonomous systems. This project has links to several research programs and teams within FCAI, such as data-efficient deep learning and autonomous systems and sustainable mobility. We seek motivated candidates with a background in computer vision or machine learning, and an interest in applying one to another.

Keywords: computer vision, deep learning

This project is supported by the Finnish Center for Artificial Intelligence his project is a part of the Finnish Center For Artificial Intelligence (FCAI). To find out more please visit https://fcai.fi/


Theoretical Frameworks and Deep Learning Algorithms for classification in Large Output Spaces

Supervisor: Rohit Babbar (Aalto University)

Research Area: Algorithms and Machine Learning

There is a growing interest in supervised learning for classification of data with large number of outputs or labels, for search and recommendation tasks. Beyond the computational challenges [1,2], due to millions of labels, in off-the-shelf usage of transformer encoders, there are statistical challenges arising from missing labels and data scarce tail-labels [3,4]. The goal of the PhD thesis would be to build efficient deep learning algorithms and study the statistical properties of the components in the learning pipeline. More information on our latest works is available here https://sites.google.com/site/rohitbabbar/Home

Requirements: Strong background in Linear Algebra, and probability/statistics. Excellent programming skills in python, and experience with deep learning frameworks (such as PyTorch).

[1] Fast Multi-Resolution Transformer Fine-tuning for Extreme Multi-label Text Classification, Neurips 2021

[2] SiameseXML: Siamese networks meet extreme classifiers with 100M labels, ICML 2021

[3] Unbiased Loss Functions for Extreme Classification With Missing Labels, https://arxiv.org/abs/2007.00237

[4] On Missing Labels, Long tails and propensities in Extreme Multi-label Classification, KDD 2022



Transformers for vision-language tasks and hyperspectral remote sensing

Supervisor: Jorma Laaksonen (Aalto University)

Researcg Area: Algorithms and Machine Learning

The project will study and develop new Transformer-based models for multimodal information processing. The application domains will include vision-language tasks, such as image and video captioning and retrieval, and modeling of hyperspectral and lidar remote sensing data. For the former, the project aims at development of new methods for describing the visual, aural and multimodal contents of video streams. Visual analyses are combined with textual content descriptions and speech recognition results for producing multimodal transcripts of all video collections. For the latter, we will analyze the information content of modern and future hyperspectral earth observation and lidar data obtained over the boreal forest zone. The results will be used in a digital twin of the Finnish forests. The project aims at novel methodological and algorithmic improvements, resulting from the need to deal with processing of huge amounts of challenging audiovisual and remote sensing data with state-of-the-art Transformer models that have already been studied and developed in the research group.



Life Science Informatics Research Projects

Deep generative modeling for continuous-time dynamics and biomedicine

Supervisor: Harri Lähdesmäki (Aalto University)

Research Area(s): Life Science Informatics (LSI); Algorithms and Machine Learning (AML)

We are looking for a doctoral student to work on deep generative modeling for applications that are at the intersection of biomedicine and health as well as continuous-time dynamics. Our research work involves developing new deep generative modeling methods for large-scale health datasets from Finnish biobanks, computational immunology, various single-cell datasets, and general data-driven modeling methods for continuous-time dynamics. Methodologically, the work revolves around e.g. conditional deep generative models, neural ODEs, and Bayesian methods. Work can focus more on method development or can also include applications, depending on applicant’s own preference. Applicants are expected to have good knowledge of machine/deep learning, statistics, programming, and (optionally) interest in bioinformatics and biomedicine. For more information, see our research group web page: https://research.cs.aalto.fi/csb/


Graph Algorithms for Long-Read Sequencing Data

Supervisor: Assoc. Prof. Alexandru Tomescu (University of Helsinki) (https://www.cs.helsinki.fi/u/tomescu/)

Research Area(s): Life Science Informatics (LSI); Algorithms and Machine Learning (AML)

Novel long-read sequencing technologies that are also extremely accurate are transforming bioinformatics analyses. The focus of this project is to develop the next generation of graph algorithms that can fully utilize such read data. We aim to (1) develop novel graph algorithms for e.g. flow decomposition problems, safety in graph problems, and (2) apply such algorithms into practical state-of-the-art bioinformatics tools (related to variants of sequence assembly).

We require a good algorithmic expertise (with a focus on graph algorithms), and strong skills for developing practical tools, running experiments on the departmental computing cluster and interpreting the results. Previous research experience and related activities such as programming contests are considered a plus.

You will be part of the Graph Algorithms Team (PI Alexandru Tomescu) of the wider Algorithmic Bioinformatics Research Group at the Department of Computer Science, University of Helsinki. Our team includes 3 PhD students, 2 Postdocs, and enjoys prestigious funding such as an European Research Council Starting Grant.


Multivariate variable selection with applications to genetics

Supervisor: Matti Pirinen (University of Helsinki)

Research Area(s): Life science informatics (LSI), Algorithms and Machine Learning (AML)

Our current ability to measure massive numbers of variables, for example, in biological sciences (millions of DNA variants) or in personal health applications (thousands of time points in activity monitoring) call for efficient ways to separate the important variables from unimportant ones. Additionally, the outcome space of the variable selection problem may be multidimensional, e.g., we may search for DNA variants that affect both measured biomarker levels and disease risk. This setting calls for methods to decompose the multivariate outcome space into most informative components.

We are looking for a PhD student to work on realistic multivariate variable selection problems where variables may have been measured incompletely and where we may have access only to statistical summaries of the underlying data. Our previous work include the Bayesian approach implemented in the FINEMAP algorithm (www.finemap.me). We have also first-hand experience on large-scale genetic data analyses, including, e.g., the world’s largest study on migraine genetics (Nature Genetics 2022, 54: 152-160).

This project includes both theoretical (statistical) and empirical (simulations and genetic data analysis) components and is connected to the most recent resources of genetic and health-related data from Finland and across the world.

Research group pages: https://www2.helsinki.fi/en/researchgroups/statistical-and-population-genetics


Networks, Networked Systems, and Services Research Projects

Energy-efficient high-performance computing ecosystems

Supervisor: Maarit Käpylä and Linh Truong (Aalto)

Research Area(s): Algorithms and Machine Learning (AML) Networks, Networked Systems, and Services (NNSS);

The energy consumption of large-scale computations and data analysis tasks is huge, but the current mindset of users and software developers for applications run on modern high-performance computing (HPC) platforms is not directed towards re-usability nor sustainability aspects. We are developing an ecosystem of software tools and services to enable HPC applications to use current and future heterogeneous HPC platforms in an optimal, energy-efficient and sustainable manner  through optimizing the HPC solutions, developing intelligent methods to avoid the data  bottleneck, harnessing the full potential of cloud computing in the analysis of huge data sets, and by integrating this into a seamless HPC ecosystem that performs the desired tasks in an automated, fault-tolerant, and reusable way. All these aspects are critical in enabling scientific applications at the Exa-scale, where Europe is heading currently with the in-auguration of the two pre-exascale supercomputers, one of which is LUMI, our target platform at CSC, Finland. We are looking for a PhD candidate with education and preferentially some experience in big data analysis platforms and their distributed machine learning solutions and/or heterogeneous distributed HPC systems.The PhD project will require contributions to both domains. 


How to Apply: Read through the list of available positions and select the topic(s) that you are interested in. Click on the link to our eRecruitment system. You will have to create an account for our eRecruitment system. However, you can use the same form to apply for multiple research areas and projects.