Stefan Engblom

Stefan Engblom

Professor in Scientific Computing

CV

See also: CV (pdf).

Research

I am particularly interested in Scientific Computing in the intersection with Data-driven research and Data Science. I have extensive experiences in many aspects of Scientific Computing in general, in Numerical Modeling and -Analys in particular, as well as to some extent in High-Performance Computing. My main focus of applications are in the Biosciences at broad, but I’ve also taken an interest in traditional computational Engineering applications.

Current active research projects include Bayesian approaches for compute intensive data-driven models in epidemics, including in particular prediction, and multiscale modeling and parameterization of living cells, where spatial stochasticity is an important aspect of the modeling.

I am currently the main supervisor for 3 PhD-students:

I was the main advisor of

I am also the secondary advisor for Anna Frigge, Alfred Andersson, and Helena Andersson. I was previously the secondary advisor of

MSc/BSc-theses:

Previously

I became an Associate Professor in 2014, being previously promoted to Docent in 2013 in Scientific computing. I originally joined UPMARC in 2011 with the aim at bringing problems from Scientific Computing into a form suitable to modern multicore/manycore computers, and vice versa, to develop and analyze algorithms and techniques suitable to such cards with interesting applications in mind. Research outputs here include, amongst others,machine learning methods in imaging with X-ray lasers, auto-tuning in CPU/GPU implementations of adaptive fast multipole methods, and shared memory approaches for event-based algorithms.

The Linnaeus center of excellence UPMARC
⇒ Focus area Application Performance
⇒ ⇒ Project group Parallel Algorithms

Before that, I was a PostDoc at the the Linné FLOW Centre where I started in September 2009 to work on computational modeling of multiphase flow for two immiscible fluids and a surface active agent. For example, this would be the correct model when considering a mixture of oil/water and a detergent.

Before that I was also briefly involved in Anna-Karin Tornberg’s project concerning simulating fibers suspended in fluids.

As a graduate student I studied methods for computing numerical solutions to stochastic descriptions of chemical reactions. The underlying mathematical description is a continuous-time Markov chain and the equation governing the probability density is called the Master Equation. Unfortunately, the master equation cannot be solved numerically for more than, say, five molecular species due to the exponential growth of work and memory requirements (‘curse of dimensionality’). Stochastic descriptions of chemical reactions are needed to describe the chemical processes taking place inside living cells with few copy numbers of each molecular species. Usual models for cell simulation are based on the reaction rate equations which form a system of nonlinear ordinary differential equations. Such models ignore the stochastic fluctuations in the cells and are therefore less accurate.

Computational systems biology group.

E. Coli meshE. Coli conc

I was also involved in a project joint with the Division for Electricity and Lightning Research, Uppsala University called “Electric power generation from winds”. The ultimate goal of the project is to provide more efficient wind turbines. I have been working together with Paul Deglaire and, lately, Anders Goude. The work has resulted in a two-dimensional random vortex method simulating fluid flows around general airfoils at a quite high Reynolds number. My contribution has been focused around a user-friendly and very efficient implementation of the fast multipole method.