Interview with Elias Jarlebring

Could you briefly describe your education and background?

I have always been interested in mathematics, computers, programming and technology in general. At the end of my undergraduate studies, I became interested in the intersection of algorithms, linear algebra and numerical analysis, which lead to the core of my current research topic "Numerical linear algebra". Germany has a very strong numerical linear algebra community, so I decided to do my PhD degree in Germany and ended up in Carl-Friedrich-Gauss-Fakultät at TU Braunschweig. Contacts I made there led me to later become a post-doc in KU Leuven, Belgium. I was awarded the Dahlquist fellowship which brought me back to KTH, and since then I have also spent longer time-periods in LBNL Berkeley, EPFL and the University of Geneva.

Could you tell us a few words about your research area?

The focus of my work is algorithms, theory and applications using linear algebra techniques. This is essentially what is meant by the topic "Numerical linear algebra". One of the most used tools for scientific computing - LAPACK, is the product of decades of research in numerical linear algebra.

Similar to many other applied mathematics topics, numerical linear algebra is fundamental in the sense that applications arise in a very wide range of fields. We often work on a mathematical problem with a specific application in mind, but phrase the result in an abstract way such that it could solve challenges in completely different fields.

Initially my work was motivated by applications in systems theory, where I already started to work with nonlinear eigenvalue problems, in particular applications in delay-differential equations. These techniques turned out to also be applicable in quantum physics and acoustics. My recent interest has included applications of the methods in data science, but also software development.

Could you tell us more about your contributions to the scientific computing language Julia, and how your research has been used there?

Some years ago, we identified that the field of nonlinear eigenvalue problems had a mature set of algorithms, but they were not used in applied fields as much as they should have. In order to improve the access of algorithms to other fields, we wanted to develop a software package. The Julia language dramatically increased in popularity because of its speed and simple syntax, so we developed it in Julia and made it a part of the Julia repository under the name NEP-PACK. I have also made contributions to the core linear algebra functionality in Julia, and the scientific machine learning package SciML. SciML has received quite some attention recently, since it is the basis of the software Pumas, one of the go-to tools used in clinical trial testing by pharmaceutical companies including Moderna.

You have recently been promoted to Professor of Numerical Analysis. What does this promotion mean to you?

The promotion to professor has been a recognition of my scientific achievements, in research, education and leadership. I am of course very grateful for this.

A professor at KTH has more leadership roles within the specific discipline but also for science in general. For me, I hope it will provide me with more opportunities to promote quality, both research and education. For example, quality enhancing topics that I enjoy working with include internationalization and mobility, respect for both fundamental and applied research, and electronic tools in education.

You have supervised several PhD students at KTH. Could you reflect on this experience?

Doing good supervision is hard and can be very rewarding. Students are different, not only in what they are good at but also in terms of personality and preferred working style. How much structured direction is suitable? How many chaotic brain-storming meetings should we have? Although I have my style, open-door policy and try to have many purely scientific discussions, I try read to and also accommodate what works well for the student.

As a junior researcher coming directly from a master's education, it is important to learn that everyone makes mistakes.

Danai Deligeorgaki