Before contacting me about employment opportunities, please read the text below!

What do we do? - the big picture

While our main "application area" is often motivated by atmospheric processes described below, our machine learning / artificial intelligence (ML/AI) research is generic and published in quality ML/AI journals and conferences. The fundamental ML/AI topics covered are probabilistic emulator / predictive regression models for atmospheric processes, randomisation methods for interactive visual data exploration, advanced statistical methods for ML/AI, and explainable AI.

We work in a multidisciplinary team of computer and atmospheric scientists. We are currently setting up our new Virtual Laboratory for Molecular Level Atmospheric Transformations (VILMA) Centre of Excellence. The objective of VILMA is to model atmospheric molecular level processes efficiently and to understand the underlying mechanisms and causal connections. VILMA will combine first-principles quantum chemical and other simulations and probabilistic ML/AI models with interactive visualisation and exploratory data analysis.

Who can we hire?

Most of our work is "AI/ML theory" that is not specific to atmospheric topics but often can be applied there. The persons we hire should advance the AI/ML research agenda described above.

Any candidate should have essential ML and related mathematics knowledge and some programming skills. We will consider applicants with computer science, mathematics, atmospheric science, physics, and chemistry backgrounds.

Interest in the atmospheric science topics described above is considered an advantage but not an absolute requirement. Prior knowledge of atmospheric processes is optional. An interest in natural sciences is an advantage.

The contract duration for doctoral students and postdoctoral researchers is agreed upon individually; the entire course of PhD studies is typically four years, and postdoctoral contracts are typically between 1 and 4 years. We welcome PhD students and postdoctoral researchers both from Finland and abroad. Doctoral students should have completed an MSc degree before starting employment, and postdoctoral researchers should have a PhD degree.

Interns and MSc students - who have, e.g., done well in my classes - will work on research-grade problems that align with our scientific interests. We primarily seek students tentatively interested in continuing to PhD studies in my group after the MSc degree; most of our PhD students have started as interns and MSc students. The contract will be agreed upon individually. Depending on the phase of studies, typical options are (i) a 3-month full-time internship project (often, but not necessarily, as a summer internship) or a part-time internship (which makes it possible to do some courses at the same time, a typical arrangement for students during teaching periods) that by mutual agreement can later continue to MSc or PhD thesis; or (ii) 6 month MSc thesis project that by mutual agreement can later continue to PhD thesis. Intern and MSc student applicants should have study right in a Finnish university. We encourage interested students from outside Finland to apply first to the relevant Bachelor's Programme or Master's Programme (e.g., data science, theoretical and computational methods, or atmospheric sciences) at the University of Helsinki.

All positions include salary with social security and pension fund payments according to Finnish legislation, Finnish national health insurance system access (details), and occupational health care services, among other staff benefits. The university covers work-related travel expenses. You can read more about practicalities from our website for arriving international staff and why you should choose the University of Helsinki.

How to contact us?

If you are interested in working with us, please email Kai Puolamäki a brief motivation letter (typically max. 1 page) where you explain why you would like to work with us and your primary interest in ML/AI. Please attach to your email a copy of your CV as well as a study transcript that displays the grades in Finnish, Swedish, or English (and clearly explains the grading scale and the maximums - an unofficial copy is ok) if you are interested in a student position or a list of publications if you are interested in a postdoctoral researcher position. I will consider the following three items when looking at your email: (i) skills and experience (demonstrated, e.g., by course grades for a student candidate or publications for a postdoctoral researcher candidate), (ii) topical fit (your skills and experience are relevant for our research plan), and (iii) motivation to work with us and interest in our research topics (demonstrated, e.g., by your motivation letter). Other documents (such as names of references, portfolio of your past work, degree certificates etc.) are not typically necessary for the first contact. Please do not send copies of your publications or thesis. We will be in touch with you if there is an opening.

Please see the VILMA job opportunities page for more generic non-AI VILMA jobs.

Examples of topics: ML/AI with real-world data

Below, you can find some of the topics that we are actively working on:

1. Explainable AI for digital twins. We use machine learning algorithms and physics simulators to model atmospheric transformations, measurement devices, and other processes. We call these models collectively "digital twins". In this project, the task is to apply methods developed, e.g., in our prior work, to find helpful and understandable explanations for these digital twins. As a starting point, you can use Björklund et al. (2019), Puolamäki et al. (2020), and/or Björklund et al. (2022).
2. Uncertainty quantification for AI. Concept drift is an issue in almost any real-world application of machine learning (atmospheric transformations included). A model trained in some circumstances (e.g., under a specific environment, to particular molecules etc.) may not work in other cases. This problem is closely related to active learning: how do you choose the training data to reduce the uncertainties most? The detection and quantification of concept drift are crucial: can we trust the outputs of our models? The other way to put it is that we want to be able to find confidence intervals for machine learning models (if the confidence intervals are wide, then there is concept drift and vice versa). You can use Oikarinen et al. (2021) or Savvides et al. (2023) as a starting point for this project.
3. Open-source tools for randomisation and exploratory data analysis. Visual exploration of high-dimensional datasets and the future of digital twins is a fundamental task in exploratory data analysis (EDA). We have developed a theoretical model for EDA, where patterns already identified and considered known by the user are input as knowledge to the exploration system. The user is shown views of the data where the user's knowledge has been considered. In this project, you will implement an open-source tool for exploratory data analysis, for example, by adding functionality to our xiplot package. The tool should be web-based, cross-platform, and scale to large datasets. Programming skills and previous experience in open-source software development are considered an advantage.

Here is a short link to this page: https://bit.ly/edahelsinki2023jobs