I am a research scientist working in the field of probabilistic modelling of complex systems at the non-profit technical safety organisation Gesellschaft für Anlagen- und Reaktorsicherheit (GRS) in Garching near Munich. Previously, I did my PhD in theoretical physics on Bayesian approaches for the investigation of complex climate time series at the Focus Area for the Dynamics of Complex Systems (DYCOS) and at the Paleoclimate Dynamics Group, both based at the University of Potsdam.
Even though the scientific disciplines - complex natural systems and complex engineered systems - are very different, my research focuses on the general understanding of stochastic transition patterns occuring and influencing all kinds of complex systems. The mutual challenge thereby may be a strikingly simple question: How can I formulate a scientific hypothesis in terms of a probabilistic approach to unravel principles guiding a complex dynamic system? Or in more elegant and ambitious words:
"If the universe is the answer, what is the question?"
From one perspective I approach this challenge by designing generic probablistic models to analyse observations of natural systems, such as the climate system. The aim is to derive probabilisitic information that helps to decipher the recorded system evolution, e.g. paleoclimate proxy records. From another perspective I approach this challenge by performing deterministic simulations traversing specific trajectories in the phase space to analyse the behavior of engineered complex dynamic systems in the presence of stochastic transition scenarios. For instance, an important study case is the behavior of a nuclear power plant given failure scenarios assigned with uncertainty. For details, please see my publications.
Common to all these approaches is the demand of high performance scientific computing and programing concepts. Therefore, I am additionally interested in information theoretical principles to implement sustainable and platform-independent (mainly pythonic) programs that are in general designed to be performed in multi-core environments.
! upcoming conference contribution !
I am looking forward to the SIAM UQ20 conference on Uncertainty Quantification in March 2020 taking place at the Technical University Munich. Here, I will talk about An agent-based approach to explore complex dynamic systems for probabilistic safety analyses and explain how the perfomance of rule-based agents can be enhanced via interactive, conditional random sampling to investigate the behavior of complex systems in the presence of critical state transitions.