Postdoctoral Researcher and Research Fellow Positions in ICT (Helsinki, Finland)

Aalto University and the University of Helsinki are looking for Postdoctoral Researchers and Research Fellows in multiple areas of ICT, including:

  • artificial intelligence and machine learning
  • data science
  • privacy and security
  • computational health
  • human-computer interaction

We welcome applications linking two or more of these areas together. We are committed to fostering an inclusive environment with people from diverse backgrounds, and researchers from underrepresented groups are particularly encouraged to apply.

The deadline for applications is January 25th, 2021 (midnight UTC+02:00). By applying to this call, organized by Helsinki Institute for Information Technology HIIT and Finnish Center for Artificial Intelligence FCAI, you apply with one application to both Aalto University and the University of Helsinki. Aalto University and the University of Helsinki are the two leading universities in Finland in computer science and information technology. Both are located in the Helsinki Metropolitan area, and the employing university will be determined according to the supervising professor. Please see below for additional information on the application process.

Helsinki Institute for Information Technology HIIT and Finnish Center for Artificial Intelligence FCAI bring together the world-class expertise of Aalto University and the University of Helsinki in joint research centers and programs. There are multiple research groups participating in each of them. The researchers to be recruited will work in one of the participating groups, or in some cases a shared position may also be possible. In addition, positions are available in the specific projects listed below.

Open positions

HIIT funded positions:

The mission of the Helsinki Institute for Information Technology HIIT is to enhance the quality, visibility and impact of Finnish research on information technology and support cooperation between ICT researchers, researchers in other fields, industry and public organisations. All excellent researchers in any area of ICT will be considered, but priority is given to candidates who support one (or more) of the strategic focus areas listed below.

  1. Artificial Intelligence. In this area our flagship unit is FCAI supporting its research programmes and highlights described above, but HIIT welcomes all excellent applications from areas of artificial intelligence research not necessarily covered by FCAI.
    Lead: Professor Samuel Kaski
  2. Data Science. Extraction of knowledge and insights from data is important across many fields of science. This research area is aligned with the Helsinki Centre for Data Science (HiDATA) aiming to create a world-class research and research-based education hub of data science in Helsinki as a joint effort between the University of Helsinki and Aalto University. HiDATA builds on the existing, strong research in various areas of data science, and aims to provide novel synergies across disciplines.
    Lead: Professor Sasu Tarkoma
  3. Cybersecurity. HIIT supports the research activities of the Helsinki-Aalto Institute for Cybersecurity (HAIC), a strategic initiative set up by Aalto University and the University of Helsinki to ensure excellence in information security research and education. The long-term mission of this research area is to enable the design, building and deployment of distributed large-scale systems, where each node (component, device, network element etc.) contributes in verifying trustworthiness of the entire system. Each node would also be able to verify whether any other node or even the entire system is trustworthy.
    Lead: Professor Valtteri Niemi
  4. Computational Health. The goal of this research area is to solve hard computational challenges faced upon the emerging digitalization and wide adoption of data-driven approaches in healthcare. We combine state-of-the-art computational methods with large-real world data arising in healthcare and personalized medicine, analysed in collaboration with experts from Aalto University, University of Helsinki, Hospital District of Helsinki and Uusimaa (HUS) as well as Institute for Molecular Medicine Finland (FIMM).
    Lead: Professor Juho Rousu
  5. Other research areas. List all areas of your expertise. HIIT has a long track record of supporting cross-disciplinary work, so feel free to include non-ICT areas of your expertise as well.

FCAI topics:

Finnish Center for Artificial Intelligence FCAI is a community of experts that brings together top talents in academia, industry and public sector to solve real-life problems using both existing and novel AI. FCAI’s research mission is to create a new type of AI that is data efficient, trustworthy, and understandable. We aim to build AI systems capable of helping their users in AI-assisted decision-making, design and modeling. We develop AI techniques needed for systems which can help their users make better decisions and design better solutions across a range of tasks from personalized medicine to materials design. A core insight in developing such AIs is that they need to have world models for understanding the world and interacting with it, and user models for understanding the user and interacting with them. Read more about our research here.

  1. Computational modeling of human motivation and experience
    One of our aims at FCAI is ‘artificial human understanding’, a type of AI that much better estimate what human collaborators want or prefer or experience. The goal is to estimate the deeper (latent) motivational processes and subjective constructs like perceived aesthetics, and thereby offer help to human collaborators more effectively. We are looking for an outstanding researcher to help us construct computational models of motivations and experience rooted in most current psychological theory, and use those models to boost the inferential capability of AI assistants. In particular, we will model designers as agents that maximize expected utility in their choices but are limited by their own bounds and that of the design tools (see Gershman et al. Science 2015). This area, called computational rationality, is an exciting convergence point of machine learning and cognitive science. The ideal candidate has demonstrated track record modeling user behavior or cognition, with publications on reinforcement learning, probabilistic models, or cognitive models.

    Previous papers:
    Roohi, Shaghayegh, et al. “Predicting Game Difficulty and Churn Without Players.” Proceedings of the Annual Symposium on Computer-Human Interaction in Play. 2020.
    Supervision: Antti Oulasvirta (Aalto University); potential other supervisors: Perttu Hämäläinen (Aalto University), Arto Klami (University of Helsinki), Elisa Mekler (Aalto University)
    Keywords: computational modeling, human motivation, computational rationality, interactive AI, reinforcement learning, cognitive models

  2. Computational cognitive models for AI assistance
    A prime goal for FCAI’s research is to develop a new form of AI that can better work with people and assist them in everyday tasks. We believe that deep integration of human cognition into the technical principles that govern the AI’s operation is necessary for ethically acceptable applications. We are now looking for an outstanding, methodologically oriented candidate to help us develop computational cognitive models based on the theory of computational rationality (Gershman 2015 Science). The ideal candidate has previous background in MDP/POMDP-based modeling, decision theory, multi-agent systems, or reinforcement learning.
    Previous papers:
    Kangasrääsiö, Antti, et al. “Parameter inference for computational cognitive models with Approximate Bayesian Computation.” Cognitive Science 43.6 (2019): e12738.
    Gebhardt, C., Oulasvirta, A., & Hilliges, O. (2020). Hierarchical Reinforcement Learning as a Model of Human Task Interleaving. arXiv preprint arXiv:2001.02122.
    Supervision: Antti Oulasvirta (Aalto University); Samuel Kaski (Aalto University)
    Keywords: Computational cognitive modeling, AI assistance, interactive AI
  3. AI-assisted modelling of dynamic interactional data
    This project is concerned with modelling on graphs where the environment is uncertain and can change, possibly due to user interaction [1]. Graph neural networks, the state-of-the-art models for embedding graphs, have almost exclusively focused on non-strategic settings. However, several important applications involve agents on networks that compete/collaborate [2] as part of decision-making. This project is concerned with modeling the uncertainties due to the different players. We are looking for post-doctoral researchers and/or doctoral students to work on the intersection of graph neural networks, Gaussian processes, and dynamical systems (see [3] for an overview). The ideal candidate will have a strong mathematical/statistical training and good programming skills.
    [1] Ruotsalo, Tuukka, Jacucci, Giulio, Myllymäki, Petri, and Kaski, Samuel. ”Interactive intent modeling: information discovery beyond search”. Communications of the ACM, 58(1):86-92, 2015.
    [2] Garg, Vikas, and Jaakkola, Tommi. “Predicting deliberative outcomes.” In International Conference on Machine Learning, pp. 3408-3418. PMLR, 2020. Link:
    [3] Solin, Arno. “Machine Learning with Signal Processing”. ICML 2020 tutorial. Link:
    Supervision: Samuel Kaski (Aalto University), Vikas Garg (Aalto University) and Arno Solin (Aalto University)
    Keywords: graph neural networks, dynamical models, Gaussian processes, stochastic differential equations, Bayesian methods
  4. Design of new materials with active and human-assisted machine learning
    One of the goals of the FCAI Research Programs is to develop AI techniques for materials design. One of the most challenging tasks of materials science is to describe the complexity of interatomic interactions on the quantum level within a coherent theoretical formalism. Recently, good progress has been made in using Gaussian regression machine learning methods to describe multi-elemental materials systems [1]. However, the large multidimensional parameter space limits the applicability of the approach to complex systems, such as the new very exciting class of materials known as high-entropy alloys. In this project, the previous approaches will be extended to include active human-guided machine-learning, to develop a data-efficient approach for customer tailored description of materials properties that maintains the quantum accuracy of the structures deemed crucial for real-life applications. The work contributes more generally to the question of how to implement human-AI collaboration in material design, aiming to develop general principles and methodology applicable beyond these specific materials and models. The candidate should have a background in quantum-level materials physics or chemistry and machine-learning approaches.
    [1] J. Byggmästar and K. Nordlund and F. Djurabekova, Gaussian approximation potentials for body-centered cubic transition metals, Phys. Rev. Materials 4, 093802 (2020)
    [2] J. Määttä, V. Bazaliy, J. Kimari, F. Djurabekova, K. Nordlund, and T. Roos, Gradient-Based Training and Pruning of Radial Basis Function Networks with an Application in Materials Physics, Neural networks 133, 123- (2021). doi: 10.1016/j.neunet.2020.10.002.
    Supervision: Flyura Djurabekova (University of Helsinki), Teemu Roos (University of Helsinki), Arto Klami (University of Helsinki)
    Keywords: Gaussian regression, Active machine learning, Materials design, interatomic interactions, high-entropy alloys, density-functional theory
  5. Sample-efficient and generalizable deep learning
    This project aims to develop learning systems capable of generalizing to new concepts and tasks from a small number of demonstrations during training. This kind of a sample-efficient generalization is a must for digital assistants that are truly cooperative and responsive, operating in heterogenous real world environments. Deep learning is a powerful learning tool which currently requires a large amount of training data and shows limited ability to generalize to out-of-distribution data. This project aims to improve sample efficiency and generalization abilities of deep learning systems. To improve sample-efficiency, we enhance the learning process by introducing relevant auxiliary tasks. In supervised learning, auxiliary tasks come from modeling unlabeled data (for example, via self-supervision or contrastive learning). In reinforcement learning, the auxiliary tasks come from building a predictive model of the environment. To improve the generalization abilities, we develop models with inductive biases that facilitate better transfer of knowledge and faster adaptation to new tasks. We develop world models that consist of a hierarchy of re-usable components (for example, hierarchical reinforcement learning) and build object-based representations of the world. The developed object-centric models should generalize well to different combinations and visual appearances of interacting objects, thus increasing the generalization ability. Object-centric models also improve the interpretability of the machine learning system as humans are used to reason about the world in terms of objects and their interactions. The developed models combine classical deep learning architectures (such as convolutional neural networks) and neural networks with the relational inductive bias (such as graph neural networks and transformers).
    Supervision: Alexander Ilin (Aalto University), Juho Kannala (Aalto University)
    Keywords: self-supervised learning, few-shot learning, meta-learning, model-based reinforcement learning, object-based representations
  6. Deep Quantum Graphical Models
    Probabilistic graphical models (PGMs) such as Markov Random Fields and Bayesian networks allow us to encode the conditional dependencies between random variables succinctly. However, these models are designed to run on the classical computer. This project is aimed at designing quantum counterparts for PGMs, and implementing them via deep architectures such as graph neural networks [1] that can be executed on a quantum computer. Another goal of the project is to investigate whether it might be possible to simulate these algorithms on a classical computer under some restrictions [2]. We are looking for post-doctoral researchers to work at the intersection of quantum machine learning, graphical models, and deep learning. The ideal candidate will have a strong mathematical inclination and training, and excellent programming skills.
    [1] Garg, Vikas, Jegelka, Stefanie, and Jaakkola, Tommi. “Generalization and Representational Limits of Graph Neural Networks”. International Conference on Machine Learning (ICML), 2020. Link:
    [2] Bondesan, Roberto and Welling, Max. “Quantum Deformed Neural Networks”. arXiv:2010.11189, 2020. Link:
    Supervision: Vikas Garg (Aalto University) and Juho Kannala (Aalto University)
    Keywords: graphical models, quantum computing, deep learning
  7. Atmospheric AI
    Artificial intelligence (AI) and machine learning (ML) are making their inroads to atmospheric and earth sciences. There are lots of opportunities to do research in physical sciences more efficiently and to obtain novel results of high impact—both in atmospheric and computer sciences—by developing and applying novel AI methods to solve scientific problems. In this project, we plan to build probabilistic models of measured and simulated natural world phenomena, trained by using simulator outputs or real-world observations, which allow us for example replace computationally expensive simulator runs with faster ML computations, to fill in missing data from observations, and to better understand complex systems and processes and underlying causal relations. Our objective is to also model the interactive data analysis and model building process of the substance area experts (here atmospheric scientists), which allows us to address problems such as how to design the exploratory data analysis workflows and systems and how to best incorporate the knowledge and insights of the experts into the model building process. We are looking for an atmospheric scientist with interest in AI, or a computer scientist who wants to develop AI methodology and work with physics-related applications. We can adjust the work plan and the supervision arrangement depending on the qualifications and interests of the hired person.
    Supervision: Kai Puolamäki (University of Helsinki); potential co-supervisors Hanna Vehkamäki (University of Helsinki), Leena Järvi (University of Helsinki), Tuomo Nieminen (University of Helsinki)
    Keywords: Atmospheric and earth sciences; exploratory data analysis; automatic experimental design; interactive user modelling; causal inference
  8. Cognitive modelling in Augmented Reality
    Augmented Reality (AR) provides numerous opportunities to gather behavioural signals for use in interactive applications. For example, the Varjo XR-1 headset includes high-definition stereoscopic video, eye-tracking and head-tracking. The goal of the project is to use reinforcement learning and cognitive modelling to understand and assist users during AR scene manipulation tasks. These tasks will involve mixed physical and virtual environments, such as interior design settings (e.g. placement of furniture) or knowledge-based tasks such as exploratory data analysis. We use reinforcement learning as a framework for modelling user interaction. In general, reinforcement learning requires substantial user feedback, which is even more challenging in the AR setting, where headsets generate massive quantities of data that need to be processed in real-time. We therefore want to develop a simulation-based AI approach to model and understand how users perceive and understand the environment presented to them by the headset, in order to provide user with a personalised and contextualised experience. The research group is already in possession of a Varjo XR-1 headset and several VR-2 pro headsets. Skills requirement: c# programming, unity; good to have knowledge of reinforcement learning, data visualisation, working with users.
    Supervision: Dorota Glowacka (University of Helsinki), Giulio Jacucci (University of Helsinki); Alan Medlar (University of Helsinki)
    Keywords: Augmented Reality, cognitive modelling, simulation-based AI, reinforcement learning
  9. Interactive reward elicitation
    A major problem in RL is to determine suitable reward functions. Reward design often requires a significant amount of trial-and-error even for experts with experience. Our core idea is to include a model of the user in the reward elicitation process such that the process can take into account the user’s limitations in addition to maximizing information gain, thus moving beyond the noisy oracle paradigm. Moreover, we plan to integrate behavioural (explicit feedback from user) as well as implicit physiological measurements.
    Supervision: Ville Kyrki (Aalto University), Simo Särkkä (Aalto University), Joni Pajarinen (Aalto University), Antti Oulasvirta (Aalto University), Samuel Kaski (Aalto University)
    Keywords: Reinforcement learning, reward design, reward elicitation, interactive AI, inverse RL
  10. Graph-based world models for sample efficient and human friendly reinforcement learning
    Reinforcement learning has shown promise in computer game play and robotics but learning long-term behavior directly from visual input has been limited to simple tasks and actual task specification has been hard for regular users. Meanwhile deep learning has been able to infer semantic information about objects and their dependencies from visual input in the form of object graphs. To make long-term planning more efficient we will use graphs as dynamic states in reinforcement learning. We will learn models of how the state of the system, that is, the graph changes on agent actions. Using the learned world model together with a reinforcement learning policy representation such as a graph neural network allows the system to generalize over different world scenes. Moreover, the approach allows for measuring closeness of the current scene to the desired one in the form of graph similarity measures. Thus, a non-expert user can specify the task objective in terms of visual scenes, for example, by uploading a set of photos of the desired end state or of states that the algorithm should avoid which are then converted to graphs. We provide good opportunities for applying the methods in mobile robotic manipulation and autonomous driving.
    Supervision: Joni Pajarinen (Aalto University), Alexander Ilin (Aalto University), Juho Kannala (Aalto University)
    Keywords: Reinforcement learning, model learning, planning, computer vision, decision-making, human feedback, robotics
  11. Planning to learning world models to plan and learn
    Planning to reach long-term goals has allowed for super-human performance in tasks with accurate models [1] while learning has allowed for solving tasks with complex inputs and outputs [2]. However, tasks with complex inputs and outputs that require long term exploration and planning are out of reach for current methods. We will go further and integrate learning and planning. We will use meta-learning [3] to learn a world model and a policy that mixes planning and reinforcement learning such that planning is used for problem parts that require principled exploration and learning is used for parts that require pattern recognition. We will develop methods that integrate learned dynamics models and planning with both model-free and model-based reinforcement learning. We expect the developed methods to enable efficient long-term decision making in high-dimensional continuous and discrete control tasks. We provide good opportunities for applying the methods in mobile robotic manipulation and autonomous driving.
    [1] Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., Bolton, A. and Chen, Y., 2017. Mastering the game of Go without human knowledge. Nature, 550(7676), pp.354-359.
    [2] Hessel, M., Modayil, J., van Hasselt, H., Schaul, T., Ostrovski, G., Dabney, W., Horgan, D., Piot, B., Azar, M.G. and Silver, D., 2018. Rainbow: Combining Improvements in Deep Reinforcement Learning. In AAAI.
    [3] Finn, C., Abbeel, P. and Levine, S., 2017. Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks. In ICML.
    Supervision: Joni Pajarinen (Aalto University), Alexander Ilin (Aalto University), Simo Särkkä (Aalto University), Ville Kyrki (Aalto University)
    Keywords: Reinforcement learning, machine learning, learning dynamics, control, planning, decision-making
  12. Transferable hierarchical reinforcement learning
    We tackle the general problem of learning to make high value decisions in dynamic systems at different time-scales. Major challenges of current approaches include high sample complexity, problem specificity of learned solutions, and opaquity of models. We will develop efficient hierarchical decision making models that allow 1) learning parts of the model separately for efficiency, 2) transferring the model to novel situations including transfer from simulation to the real world, and 3) demonstrating model parts and their interplay to users to bridge the human-computer gap. Contrary to prior work that does not learn models or learns only low level models we learn models on all levels allowing for quick transfer to new tasks and interpretability for users. We are looking for a researcher interested in developing the new decision making models and methods, with options on applying the techniques to autonomous driving and robotics.
    Supervision: Joni Pajarinen (Aalto University), Alexander Ilin (Aalto University)
    Keywords: Reinforcement learning, learning dynamics, control, machine learning, planning, decision making under uncertainty
  13. Semi-model-based learning of dynamics and cost functions in stochastic control
    We consider partially observed planning/stochastic control problems with continuous state spaces and with partially unknown dynamics and cost functions. The aim is to learn the dynamic model and cost functions leading to given decision behaviour while taking prior information on the dynamic and measurement models into account (see [1,2,3]). The aim is to apply this to both animal behavioural estimation as well as to human subjects. The aim would also be to identify the cost function related to interaction between the computer (AI) system and the subject. We can do mouse and brain imaging experiments, where we change the goal of the task based on previous decisions. Other applications in the field of robotics and autonomous systems can be found in active imitation learning [4].
    [1] Mombaur, Katja, Anh Truong, and Jean-Paul Laumond. “From human to humanoid locomotion—an inverse optimal control approach.” Autonomous robots 28.3 (2010): 369-383.
    [2] Simo Särkkä, Mauricio A. Álvarez, Neil D. Lawrence (2019). Gaussian Process Latent Force Models for Learning and Stochastic Control of Physical Systems. IEEE Transactions on Automatic Control, Volume 64, Issue 7, Pages 2953-2960.
    [3] Simo Särkkä (2020). The Use of Gaussian Processes in System Identification. To appear in Encyclopedia of systems and control, 2nd edition.
    [4] Hussein, Ahmed, et al. “Imitation learning: A survey of learning methods.” ACM Computing Surveys (CSUR) 50.2 (2017): 1-35.
    Supervision: Simo Särkkä (Aalto University), Joni Pajarinen (Aalto University), Ville Kyrki (Aalto University)
    Keywords: Stochastic control, reinforcement learning, machine learning, inverse optimal control
  14. Context-based curriculum learning for safe exploration
    Contrary to state-of-the-art curriculum learning that incrementally adapts the underlying task to make learning more efficient [1], we will use curriculum learning to also improve safety in exploration on real systems. We will use context based curriculum learning where the agent controls the context. By controlling the context the agent can explore different behaviors without risking the system or users. In human-robot interaction, the context can be defined in terms of the user, for example, user pose or pose of an object the user is holding. To change the context, the robot asks the human to execute the desired pose or hand out a specific object. With this approach we expect to be able to learn behavior safely on challenging physical systems. Technically the approach will require probabilistic modeling of a safety index (for example, safety margin distance between user and robot) that is learned interactively with a user that controls the context. Then exploration in reinforcement learning will need to integrate the safety index such that a minimum required safety is guaranteed, which can be achieved by controlling the learning curriculum.
    [1] Klink, P., D’Eramo, C., Peters, J. and Pajarinen, J., 2020. Self-Paced Deep Reinforcement Learning. Advances in Neural Information Processing Systems (NeurIPS)
    Supervision: Joni Pajarinen (Aalto University), Ville Kyrki (Aalto)
    Keywords: Reinforcement learning, curriculum learning, safety, planning, decision making under uncertainty
  15. Large-scale active simultaneous localization and mapping
    The aim of this task is to enable simultaneous localization and mapping (SLAM) algorithms for adapting the automated systems when the environment varies and human interaction is present. The algorithms are partly based on neural networks which analyse validity of the mapping data and execute updates, when more reliable observations are present. Learned maps can also be divided into different layers according to their dynamic nature and expected changing frequency (falling leaves, day and night, snowstorms, etc.). By leveraging the diverse field of SLAM research, the goal in this topic is to develop SLAM methods that take human aspects into account. The first level of this is to elicitate knowledge of the human preference of movement and changes in the environment, improving data efficiency and possible safety aspects of world perception. The second level is related to human-assisted SLAM under uncertainty, where the AI learns to perceive the world under multiple interleaved goals and input from the human actors co-operating with the AI.
    Supervision: Arno Solin (Aalto University), Juho Kannala (Aalto University), Matti Kutila (VTT), Laura Ruotsalainen (University of Helsinki), Simo Särkkä (Aalto University), Ville Kyrki (Aalto University)
    Keywords: Active learning, decision-making, sequential design of experiments
  16. Data efficient learning for fully autonomous robots
    This project aims to advance the methodologies that are essential to the functioning of fully autonomous robots. Primarily, they must be able to learn behaviours with minimal human effort. Reinforcement learning (RL) algorithms deal with this problem but their practical impact is limited by their data-efficiency. This could potentially be improved by learning dynamics models of the environment and utilizing it for planning (model-based RL). While past works on online trajectory optimization with simulators have demonstrated robust control in high-dimensional control problems (such as control of a humanoid), achieving the same with learned dynamics models have proven challenging. In this project, we aim to scale this model-based RL approach to high-dimensional observation and action spaces, in a data-efficient manner. We will investigate self-supervised learning of state representations based on geometric correspondence and object-centric representations, to deal with high-dimensional observations. To deal with high-dimensional action spaces, we will investigate gradient-based planning with deep neural network dynamics models and develop regularization methods to account for predictive uncertainties during planning. We also take a holistic approach where these components are studied together and evaluated on real-world robot tasks.
    Supervision: Juho Kannala (Aalto University), Alexander Ilin (Aalto University), Joni Pajarinen (Aalto University), Petteri Nurmi (University of Helsinki)
    Keywords: autonomous control, autonomous decision making, model-based reinforcement learning, machine perception
  17. Robust radio signal-based navigation
    Availability and reliability of global navigation and timing solution provided by Global Navigation Satellite Systems (GNSS) is crucial for many critical applications, such as autonomous systems. Other radio signals, such as 5G, radar and Wi-Fi may be used for complementing or replacing GNSS signals in case they are degraded or denied. This task will develop novel methods for overcoming the challenges in labeling the radio signal data, such as active learning. The task will also develop novel deep learning based methods for detecting signal abnormalities, mitigating their effects and fusing signals with each other or with other sensor measurements for robust navigation and timing results. Since the signals are heavily affected by the operational environment (e.g. being indoors affects the GNSS signal quite similarly as intentional signal interference), interaction with the user is essential. Therefore, our task will develop Reinforcement Learning based methods for selecting the best mitigation methods against signal degradation.
    [1] Mäkelä, M., Rantanen, J., Ilinca, J., Kirkko-Jaakkola, M., Kaasalainen, S., Ruotsalainen, L. (2020). Cooperative Environment Recognition Utilizing UWB Waveforms and CNNs. In Proceedings of The European Navigation Conference, IEEE, accepted.
    [2] Ferrara, NG., Bhuiyan, MZH., Söderholm, S., Ruotsalainen, L., and H Kuusniemi (2018). A new implementation of narrowband interference detection, characterization, and mitigation technique for a software-defined multi-GNSS receiver, GPS Solutions 22 (4).
    Supervision: Laura Ruotsalainen (University of Helsinki), Arno Solin (Aalto University), Matti Kutila (VTT), Petteri Nurmi (University of Helsinki), Dorota Glowacka (University of Helsinki)
    Keywords: radio signals, active learning, convolutional recurrent neural networks
  18. Data augmentation, noise and active learning — A Bayesian brain approach
    Data augmentation is crucial for modern deep learning methods. However, it is usually done in a very heuristic and manually designed way. Recently, the possibility of learning to do such data augmentation from data, especially in a probabilistic framework, has received increasing attention. Interestingly, something akin to data augmentation occurs naturally in biological brains, which rely on noisy sensory inputs and noisy neural circuits. Biological systems learn to push such noise in directions that leave the underlying inference invariant, effectively learning the local equivariance structure of the inputs. Another important strategy adopted by biological systems is active learning. Here, our goal is to develop probabilistic models which enable learning of data augmentation, in particular by drawing inspiration from Bayesian approaches to brain function, and exploit such probabilistic representations to perform active learning. This project aims at making existing machine learning algorithms more efficient while at the same time elucidating deep connections between data augmentation, noise and Bayesian active learning in neural networks, both artificial and biological. The candidate should have a solid background in probabilistic modelling.
    Supervision: Aapo Hyvärinen (University of Helsinki) and Luigi Acerbi (University of Helsinki)
    Keywords: Probabilistic modelling, Bayesian brain, Data augmentation, Active learning
  19. Bayesian deep active learning for amortized inference of simulator models
    Recent approaches to inference in simulator models exploit the power of flexible deep neural density estimators to iteratively learn a direct mapping from summary statistics of the data to the posterior distribution, skipping the intermediate steps of approximate inference. In some limited cases, the trained networks can be immediately used on new data, achieving the holy grail of amortized Bayesian inference — inference at virtually no cost at runtime. However, training these networks requires a very large number of samples from the model, and the mapping to the posterior has no notion of uncertainty, meaning that the network could fail silently in unseen regions of parameter space. This project is concerned with applying Bayesian principles of uncertainty estimation and active learning to develop a new generation of algorithms for sample-efficient training of robust, safe emulator networks for simulator-based inference. The ideal candidate has prior experience with deep learning and Bayesian methods.
    Supervision: Luigi Acerbi (University of Helsinki); Jukka Corander (University of Helsinki), Samuel Kaski (Aalto University)
    Keywords: Simulator-based inference; Bayesian deep learning; active learning; neural density estimators; amortized inference
  20. Fast active-sampling approximate Bayesian inference for everyone
    In recent years, a new approach to approximate Bayesian inference has emerged, on the side of the traditional workhorses (MCMC and variational inference). Active-sampling Bayesian inference aims to build posterior distributions (and approximations of the model evidence) in a sample-efficient way, by constructing a statistical surrogate of the posterior (or likelihood), such as via a Gaussian process, and then actively evaluating the log-likelihood or log-joint distribution where needed to efficiently update the surrogate model [1-3]. This approach is similar to Bayesian optimization, but the goal differs in that the goal is to learn the posterior distribution (and/or the marginal likelihood). Crucially, thanks to recent advances in Bayesian nonparametrics, this approach is not limited anymore to “expensive” models, but it could become part of the standard Bayesian workflow for many models, affording calculation of cheap, uncertainty-aware posterior approximations with only a small number of evaluations. This project is concerned with pushing the state-of-the-art of active-sampling Bayesian inference algorithms, in terms of both theory and implementation, to obtain a new instrument for approximate inference which would be widely accessible, fast and failsafe. The ideal candidate has prior experience with Gaussian processes and active learning (e.g., Bayesian optimization), both in theory and with modern software implementations (e.g., GPyTorch).
    [1] Acerbi L (2018). Variational Bayesian Monte Carlo, NeurIPS.
    [2] Järvenpää M., Gutmann MU, Vehtari A., and Marttinen P (2020). Parallel Gaussian process surrogate Bayesian inference with noisy likelihood evaluations. Bayesian Analysis.
    [3] Acerbi L (2020). Variational Bayesian Monte Carlo with Noisy Likelihoods, NeurIPS.
    Supervision: Luigi Acerbi (University of Helsinki); Aki Vehtari (Aalto University), Samuel Kaski (Aalto University), Arto Klami (University of Helsinki)
    Keywords: Bayesian inference; active learning; Gaussian processes; Bayesian optimization
  21. Prior constraints in probabilistic programming
    Using prior domain knowledge on model parameters and transformations is at the core of Bayesian modeling, helping to build more interpretable models that can be estimated from less data. This project develops richer ways of encoding prior knowledge, focusing on incorporating (soft and hard) constraints into probabilistic programs. The current tools support simple constraints (non-negativity of a parameters, or linearity of a function) but often the prior knowledge is in form of more complex constraints (e.g. monotonicity or near-linearity of a function, or permutation invariance) that remain challenging. Building on existing theoretical foundations for specific cases, you will work on developing both the theory and practical inference algorithms for handling such constraints.
    The position is ideal for candidates with strong background in Bayesian modeling or machine learning. Your main task is to develop the required computational methods and ideally proceed to implement them into existing probabilistic programming tools in collaboration with others. We already have several concrete applications with such prior knowledge (e.g. physical knowledge in material design and cognitive theories in decision-making), and you will work in collaboration with other FCAI projects to apply the methods in selected interesting use-cases.
    Supervision: Arto Klami (University of Helsinki); Aki Vehtari (Aalto University)
    Keywords: Probabilistic programming, Bayesian modeling, Prior knowledge
  22. Visualization in modeling workflow
    We develop AI techniques needed for systems which can help their users make better decisions and design better solutions across a range of tasks from personalized medicine to materials design. A core insight in developing such AIs is that they need to have world models for understanding the world and interacting with it, and user models for understanding the user and interacting with them. Many parts of the probabilistic modeling workflow benefit from visualization. This project develops tools for AI-assisted visualizations using AI which has a theory of mind of the user. The work will be built on existing theory in cognitive sciences and human-computer interaction. The goal is to generate visually appealing, task-specific, and informative visualizations with controllable complexity depending on the amount of information that is available, required, and sensible given the expertise of the user.
    Supervision: Aki Vehtari (Aalto University); Antti Oulasvirta (Aalto University); Arto Klami (University of Helsinki)
    Keywords: Interactive probabilistic modeling, modeling workflow, visualization, uncertainty quantification, decision-making
  23. Computer assisted Bayesian workflow
    Probabilistic programming languages make it easier to specify and fit Bayesian models, but this still leaves us with many options regarding constructing, evaluating, and using these models, along with many remaining challenges in computation. Using Bayesian inference to solve real-world problems requires not only statistical skills, subject matter knowledge, and programming, but also awareness of the decisions made in the process of data analysis. All of these aspects can be understood as part of a tangled workflow of applied Bayesian statistics. Beyond inference, the workflow also includes iterative model building, model checking, validation and troubleshooting of computational problems, model understanding, and model comparison. The goal is to develop for different parts of the workflow self-diagnosing tools that can be used as part of interactive computer assisted workflow for probabilistic model building and Bayesian data analysis.
    [1] Gelman, Vehtari, Simpson et al (2020). Bayesian workflow.
    Supervision: Aki Vehtari (Aalto University); Arto Klami (University of Helsinki); Antti Oulasvirta (Aalto University)
    Keywords: probabilistic modeling workflow, Bayesian workflow, diagnostics
  24. AI methods for neuroimaging (in partnership with DataIA)
    This project develops AI methods for analysing neuroimaging data. Such analysis will help fight brain disorders, which are increasingly seen as the biggest health challenge of the 21st century. Neuroimaging applications we consider include: 1) AI methods optimizing data efficiency will enable early diagnosis of slowly progressing brain disorders, such as Alzheimer’s, which greatly improves therapeutic outcomes; 2) Deep learning analysis of sleep patterns will enable detailed analysis of sleep disorders; 3) AI methods using causality will find where epileptic seizures originate, thus enabling precise surgical interventions; 4) Brain-computer interfaces will bring the human patient into the loop, and enable, for example, efficient rehabilitation of people suffering from strokes; 5) Studying brain responses to complex stimuli, such as video or stories, across populations enables characterizations of individual variability; 6) AI methods to solve non-linear inverse problems in neurobiology. This project is in collaboration with our partner institute DataIA located at Université Paris-Saclay, the most highly ranked university in the EU. We are looking for a candidate with either a) a PhD in a quantitative field such as machine learning, statistics or signal processing, or b) a PhD in neuroimaging with a strong computational component. The postdoc will work towards a subset of the above-mentioned topics, depending on the competences and interests of the candidate.
    Supervision: Aapo Hyvärinen (University of Helsinki), Lauri Parkkonen (Aalto University), Alex Gramfort (DataIA), Bertrand Thirion (DataIA)
    Keywords: Neuroimaging, neuroscience, deep learning, transfer learning, learning with few samples, AI-assisted diagnosis
  25. Probabilistic multi-agent modelling for AI assistance
    A prime goal for FCAI’s research is to develop a new form of AI that can better work with people and assist them in everyday tasks. This can be seen as a probabilistic modelling task which requires data-efficient inference on multi-agent models, and prior knowledge from cognitive science. We are now looking for an outstanding machine learning researcher who wants to develop with us the theory and inference methods for this new task. This will involve multi-agent modelling, POMDPs and reinforcement learning, and inverse reinforcement learning.
    Previous papers: Tomi Peltola, Mustafa Mert Çelikok, Pedram Daee, Samuel Kaski (2019). Machine Teaching of Active Sequential Learners Conference on Neural Information Processing Systems, NeurIPS 2019
    Supervision: Samuel Kaski (Aalto University); Antti Oulasvirta (Aalto University)
    Keywords: Multi-agent modelling, reinforcement learning, computational cognitive modeling, AI assistance, interactive AI
  26. AI-assisted modelling
    Modelling is a combination of a design task of building the model, and the statistical task of fitting the model to data or more generally statistical inference. While probabilistic programming is progressing fast on the latter part, and AutoML helps when there is enough data, much less help exists for the design task. We formulate the task of offering design help as a broader modelling task, which includes the modeller in the loop. The solution of the broader task gives both the AI-assistance and a solution of the primary task.
    Supervision: Samuel Kaski (Aalto University), Aki Vehtari (Aalto University), Arto Klami (University of Helsinki), Antti Oulasvirta (Aalto University)
    Keywords: Sequential design of experiments, prior elicitation
  27. AI-assisted design of experiments and interventions
    We develop probabilistic modelling and inference techniques for designing the design-build-test-learn cycles which are ubiquitous in engineering systems, and experimental design in sciences and medicine. This needs Bayesian experimental design techniques able to work well with both simulators, measurement data and humans in the loop, who are both information sources and the final decision makers. Furthermore, we need automatic design of interventions (actions) for learning causal models partially from a combination of observational and interventional data.
    We are looking for a probabilistic modelling researcher interested in developing the new methods, with options on applying the techniques to improve modelling in highlight applications of FCAI, and synthetic biology, interface design and medicine.
    Supervision: Samuel Kaski (Aalto University); Ville Kyrki (Aalto University); Kai Puolamäki (University of Helsinki); Antti Oulasvirta (Aalto University)
    Keywords: Sequential design of experiments, Bayesian experimental design, likelihood-free inference, implicit models, Causal inference

PI funded positions:

These projects are not directly funded by HIIT or FCAI, but by individual research groups.

  1. ML engineering
    We are looking for postdocs and doctoral students to work on tools and methodologies for the software engineering of machine learning systems. To ensure that machine learning systems work for real, new ways are needed to ensure their correct and efficient operation as well as their smooth development, monitoring, and maintenance. At the moment we are running or about to start multiple European and national projects focusing on testing of AI systems, MLOps, and big data analytics.
    The aim of our empirical and experimental approach is to come up with new and improved solutions for ML system development and operation. The candidate is expected to analyze, measure, and model alternative approaches and create new ideas and insights based on those. This includes implementing research prototypes to try out ideas and to collect and analyze data. Documenting the results in scientific papers is naturally important. In addition to research, the Postdoctoral Researcher is expected to contribute to other common academic tasks such as teaching, generation of new research ideas, and writing grant proposals.
    Supervision: Jukka K Nurminen ( and Tommi Mikkonen ( (University of Helsinki)

  2. Multi-omics machine learning for precision oncology
    The research group of Prof. Tero Aittokallio focuses on developing integrated computational-experimental approaches that combine multi-omics genomic profiling and clinical information from cancer patients using mathematical and statistical approaches such as machine learning and network modelling. The group is now seeking a motivated postdoctoral researcher, with strong computational and analytical skills, and an interest in applying computational and systems biology approaches for analysing, modelling and integration of large-scale molecular profiling and functional datasets. The researcher will be working in a multi-disciplinary project, funded by EU/ERA PerMed, where our objective is to integrate genome-wide profiles of cancer patients with their clinical information, with the aim of finding individualized treatment options based on response-predictive biomarkers.
    Supervision: Tero Aittokallio (Institute for Molecular Medicine Finland, FIMM, HiLIFE, FCHealth, HIIT).

  3. Next-generation databases powered with machine learning techniques
    We are looking for a postdoc researcher to develop methods for next-generation database systems powered by machine learning methods. Applicants are expected to have a background in database and/or machine learning research. Group website:
    Supervision: Jiaheng Lu ( (University of Helsinki)

  4. Machine learning for ultrasonics
    Ultrasound is a powerful tool for monitoring and cleaning structures. Direct physical modelling of ultrasound propagation enables, for example, detecting fouling and focusing high power for cleaning it, but this is only feasible for extremely simple structures and in laboratory conditions. For more complex environments we need physically motivated machine learning methods for ultrasound propagation, which we develop based on Gaussian processes and other Bayesian models, as well as deep learning. We have already demonstrated accurate fouling detection without supervision and are working towards automated monitoring of complex industrial environments with AI-controlled smart sensor network, to be eventually demonstrated in real environments with an industrial partner.
    We are looking for 1-2 postdoctoral researchers to join the interdisciplinary team combining computer science (Arto Klami) and theoretical and experimental ultrasound physics (Edward Haeggström and Ari Salmi). We are looking for candidates for the machine learning side of the project, with background in probabilistic modelling, signal processing and/or physics. The specific tasks will be determined based on interests and qualifications of the candidate. For example, you could focus on analysis and data fusion methods for high frequency signals, fundamental questions on incorporating physical knowledge in probabilistic models, or reinforcement learning for real-time control of laser-induced ultrasonics.
    Supervision: Arto Klami (University of Helsinki)

  5. Mobile Cross Reality through Immersive Computing (MeXICO)
    The MeXICO project aims to overcome the limited resources of mobile devices for novel applications in cross-reality (XR): virtual, augmented, and mixed reality. Current solutions for interactive and mobile XR require real-time rendering but they are constrained by the limited resources of mobile devices. A promising approach to overcome this issue is to offload most of the heavy computing tasks from the mobile device to a remote processing unit in the cloud or at the edge. Unfortunately, this approach faces challenges in terms of both latency and bandwidth. In fact, a noticeable motion-to-photon latency is highly detrimental in interactive applications, as it may even cause severe discomfort in addition to a poor user experience. Moreover, transmitting high-quality graphic content from remote servers to mobile devices requires a large amount of network bandwidth. Motivated by these challenges, MeXICO explores solutions for mobile and distributed XR to enable novel and effective applications.
    Supervision: Mario Di Francesco, Matti Siekkinen (Aalto University)

  6. Probabilistic modelling and Bayesian machine learning
    I am looking for a postdoc or research fellow to join the Aalto Probabilistic Machine Learning Group, to work on new probabilistic models and inference techniques. I am open to excellent and exciting suggestions, in particular around the following topics: (1) simulator-based inference, for combining first-principles models with learning from data, (2) Bayesian deep learning, (3) Bayesian reinforcement learning and inverse reinforcement learning, and (4) multi-agent modelling, and (4) privacy-preserving machine learning and synthetic data generation. Can be theoretical or applied work or both; the group has excellent opportunities for collaboration with topnotch partners in multiple applications, including user interaction, health, genomics, and neuroscience. Links:
    Supervision: Samuel Kaski (

  7. Probabilistic modelling for health, genomics, and neuroscience
    I am looking for a postdoc or research fellow to join us in developing new probabilistic modelling and machine learning methods needed in the core problems of modern healthcare: developing better drugs and personalizing treatments, based on clinical, genomic and (functional) brain imaging data. We combine the ability of modern flexible models to take into account nonlinearities and interactions, with Bayesian inference to handle the uncertainty in the data and results. Precision medicine needs causal inference and predictive modelling based on genomic and clinical data, and drug development additionally generative models of chemistry; both need adaptive experimental design. This is an excellent opportunity to work with top-notch experts in both medicine (cancer and clinical) and machine learning. Link:
    Supervision: Samuel Kaski (

  8. Bayesian deep learning
    We are looking for a researcher to join the Aalto Probabilistic Machine Learning Group, to work on developing state-of-the-art Bayesian deep learning. Key research questions are about more useful neural parameterisations, process priors on function spaces and more efficient probabilistic inference methods for deep neural networks. Possible applications range from large-scale image classification to sample-efficient Bayesian reinforcement learning and robotics.
    This work will build on top of existing research lines in the group on RL and BNNs, with a recent highlight work of implicit BNNs with state-of-the-art ImageNet performance while maintaining Bayesian principles. The group has excellent collaboration and application opportunities. Background in machine learning, statistics or math is expected.
    Supervision: Samuel Kaski (, Markus Heinonen (

  9. Privacy-preserving synthetic data generation
    Access to high quality data has become a requirement of modern data-driven techniques commonly used in research and practical applications. However, long-standing concerns and recent regulations surrounding privacy make it difficult to publish data, preventing some fields, most notably (personalized) medicine, from fully benefiting from these methods. We are looking for motivated doctoral students to join the Aalto Probabilistic Machine Learning Group in further developing our approach of creating synthetic twins of such sensitive data via probabilistic models that simultaneously aim to maximise the information retained from the original data as well as ensuring strict formal notions of privacy. We are especially eager to welcome you if you bring existing knowledge of differential privacy, generative models and probabilistic modelling.
    Supervision: Samuel Kaski (

  10. HAIC: Open post-doc position in Prof. Janne Lindqvist’s group – security engineering and usable security
    We are looking for post-docs interested in security engineering, usable security and human-computer interaction. Background and interest in systems security, security engineering, data science, machine learning, modeling, human-computer interaction or social and behavioral sciences is required. Exact topic will be agreed together with the applicant. Examples of work done in the group can be found at the website for the group The post-docs will also get to participate in the activities of HAIC. Please contact me at the email address about these positions.
    Janne Lindqvist (Aalto University)

  11. HAIC: Open post-doc position in Prof. Janne Lindqvist’s group – artificial intelligence and machine learning for systems security and privacy
    We are looking for post-docs interested in developing novel artificial intelligence and machine learning approaches to security engineering and systems security and privacy.. Background and interest in data science, machine learning, statistics and computational approaches to computer science are required. Examples of work done in the group can be found at the website for the group The post-docs will also get to participate in the activities of HAIC. Please see specific examples also Please contact me at the email address about these positions.
    Janne Lindqvist (Aalto University)

  12. Open post-doc position in Prof. Janne Lindqvist’s group – multitasking and productivity tools
    We are looking for post-docs interested in understanding productivity tools and multitasking. Background and interest in measuring user experience, modeling, human-computer interaction, computer science or social and behavioral sciences is required. Examples of work done in the group can be found at the website for the group Please contact me at the email address about these positions.
    Janne Lindqvist (Aalto University)

  13. Open post-doc position in Prof. Janne Lindqvist’s group – mixed methods HCI and security research
    We are looking for post-docs interested in pushing the envelope in mixed methods HCI and security research. Background and interest in measuring either qualitative methods or quantitative methods, and interested to learning new methods, user experience, modeling, human-computer interaction, computer science or social and behavioral sciences is required. Examples of work done in the group can be found at the website for the group Please contact me at the email address about these positions.
    Janne Lindqvist (Aalto University)

  14. Deep learning for electronic health records
    We will develop novel deep learning models for healthcare time series data. This work is done in collaboration with the Finnish Institute for Health and Welfare and other healthcare data holder in the Helsinki region. In particular, we will focus on 1) probabilistic deep learning, to address uncertainty in the predictions, 2) causal inference, required for decision making on individual and population levels, and 3) interpretability, which helps communicate the results to patients and policy makers. Also important are topics related to privacy and anonymization. The advances made in the project are central for the trustworthiness and acceptability of the methods in practice.
    Supervision: Pekka Marttinen (

  15. Reconstructing Crisis Narratives for Trustworthy Communication and Cooperative Agency
    The research project, jointly conducted between Aalto University and the Finnish Institute for Health and Welfare (THL), proposes 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. We are seeking researchers to work at the intersection of Human-Computer Interaction (HCI), design research, computational social sciences, and public health for critical societal impact. They will help conduct 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. They will also develop tools for visualizing crisis narratives to support understanding and collaborative sensemaking among key stakeholders and diverse publics. 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. We expect the candidates to have backgrounds in computer science, media and communication studies, social science, or similar disciplines. Project website.
    Supervision: Nitin Sawhney (Aalto University), potential other supervisors: Teemu Leinonen (Aalto University), Nelimarkka, Matt (University of Helsinki), Sivelä Jonas (THL)
    Keywords: Crisis Informatics, Computational Social Science, Machine Learning, Conversational Analysis, Information Visualization, Qualitative Research, Participatory Design

  16. Combining generative modeling and physical simulation for next-generation computer graphics and vision
    Computer vision and graphics meet at a common point of pain: the model of scene geometry and appearance. To yield photorealistic results, graphics algorithms require an essentially perfect geometry and apperance model. Yet, the capability of computer vision algorithms to robustly and accurately reason about the 3D shape and appearance of the world, unfortunately, greatly lags behind the capabilities to detect, recognize, segment, and so on. A great discrepancy exists between the semantic and the pixel-perfect, accurate shape and appearance. Bridging this chasm is the goal of this research.
    This entails solving long-standing unsolved problems in computer vision through the aid of computer graphics and machine learning. First, we seek to simultaneously capture accurate 3D shape and appearance of complex real-world scenes from photographic inputs; second, we seek to extend these capabilities still further to“zero-shot” generative modelling. These extremely ambitious goals will be reached by marrying simulation (rendering) and machine learning, building on the PI’s leading expertise on (1) physically-based image synthesis, (2) learned generative modeling of photorealistic images through deep convolutional neural networks (GANs), (3) capture photorealistic material appearance models using commodity devices.
    The ideal candidate will have a PhD and a strong background in deep learning, computer vision, and/or physically-based rendering.
    Jaakko Lehtinen (, Aalto University

  17. Deep generative modeling for biomedicine and genomics

    We are looking for post-docs to work on deep generative and probabilistic modeling for biomedical and genomics applications. Our research group is currently working on several important biomedical challenges, such as (i) analysis of longitudinal health data from large-scale Finnish biobanks, (ii) personalized prediction of immunotherapy efficiency for cancer patients using modern single-cell genomics data, and (iii) analysis and design of protein sequences for therapeutic applications. We are developing novel deep generative modeling methods to analyse aforementioned health data sets. Depending on the application, deep learning methods include e.g. variational auto-encoders for longitudinal/time-series data, sequence models (LSTMs/transformers) for modeling and designing proteins, Bayesian (deep) neural networks, and Gaussian processes. Post-doc can work on one (or several) of these topics, depending on the preference. Work can focus solely on developing novel methods or also apply them to exciting real-world data from our national or international collaborators. For more information and relevant recent work, see or contact Harri Lähdesmäki (
    Supervision: Harri Lähdesmäki (

How to apply

The applications are to be submitted through the eRecruitment system. Choose in the application form one or more of the research centers, programs and/or projects described above and explain in the motivation letter how you could contribute in the selected research area(s).

Required attachments:

  • Cover letter which includes your motivation for applying for the specific position(s), how you could contribute to them, and other relevant information (no more than 2 pages). In case you have selected topic “5. Other research areas”, please describe all areas of your expertise. HIIT has a long track record of supporting cross-disciplinary work, so feel free to include also non-ICT areas of your expertise that may be useful in multidisciplinary collaboration.
  • CV
  • List of publications (please do not attach full copies of publications)
  • A transcript of doctoral studies and the degree certificate of the doctoral degree

All material should be submitted in English. The application materials will not be returned. Short-listed candidates may be invited for an interview either at the Otaniemi or Kumpula campus or for an interview conducted via Skype.


The candidate should have a PhD and is expected to have an excellent track record in scientific research in one or more fields relevant to the position. Good command of English is a necessary prerequisite. In the review process, particular emphasis is put on the quality of the candidate’s previous research and international experience, together with the substance, innovativeness, and feasibility of the research plan, and its relevance to the research group or groups in question. Efficient and successful completion of studies is considered an additional merit.

Compensation, working hours and place of work

The salary for a postdoctoral researcher starts typically from about 3500 EUR per month depending on experience and qualifications. In addition to the salary, the contract includes occupational health benefits, and Finland has a comprehensive social security system. The annual total workload of teaching staff at the recruiting universities is 1624 hours.

The position is located at Aalto University’s Otaniemi campus or the University of Helsinki’s Kumpula campus. The positions belong to the university career system and the selected persons will be appointed for fixed-term positions, for postdoctoral researchers typically for two years with an option for renewal. For exceptional candidates, a longer term Research Fellow position can be considered. The length of the contract and starting and ending dates are negotiable. In addition to research work, the persons hired are expected to participate in the supervision of students and teaching following the standard practices of the hiring department.

About Helsinki

The Helsinki Metropolitan area forms a world-class information technology hub, attracting leading scientists and researchers in various fields of ICT and related disciplines. Moreover, as the birth place of Linux, and the home base of Nokia/Alcatel-Lucent/Bell Labs, F-Secure, Rovio, Supercell, Slush and numerous other technologies and innovations, Helsinki is becoming one of the leading technology startup hubs in Europe. Finland tops international rankings in education, equality and safety, and Helsinki, the capital of Finland, is regularly ranked as one of the most livable cities in the world.

About the host institutions

Aalto University is a community of bold thinkers where science and art meet technology and business. We are committed to high-quality research with significant impact on the international scientific community, industry and business, as well as the society at large. Aalto is an international community: more than 30% of our academic personnel are non-Finns. Aalto University has six schools with nearly 20 000 students; it is in world’s top-10 of young universities (QS Top 50 under 50).
For more information, see

The University of Helsinki, established in 1640, is one of the world’s leading universities for multidisciplinary research. The University of Helsinki is an international academic community of 40,000 students and staff members. The university lays special emphasis on the quality of education and research, and it is a member of the League of the European Research Universities (LERU). For more information, see

Helsinki Institute for Information Technology HIIT is a joint research institute of Aalto University and the University of Helsinki for basic and applied research on information technology. HIIT’s mission is to conduct top-level research, seamlessly moving between fundamental methods and technologies to novel applications and their impact on people and society. For more information, see

Finnish Center for Artificial Intelligence FCAI brings together the world-class expertise of Aalto University and the University of Helsinki in AI research, strengthened further with an extensive set of companies and public sector partners. FCAI has been selected as one of the prestigious Flagships of the Academy of Finland, a status granted to very few selected centers of excellence with high societal impact. The total budget of FCAI is 250 M€ over the next 8 years. For more information see

Further information

  • Research related questions: The leader of the research center/program/project specified above
  • Application process and practicalities: Sanni Kirmanen (