What cutting-edge project are you working on right now?

We work on optimal engineering decision-making under uncertainty, focusing on applications involving risk due to extreme events. For example, how to efficiently maintain infrastructure to prevent failures? Or how to plan flood mitigation to balance risk and cost? To answer such questions, we develop methods for reliability and risk assessment, sensitivity analysis, stochastic optimization and sequential decision processes. We focus on methods that can handle large uncertainty in complex systems. To give one specific example: We develop sensitivity analysis metrics for predictive modeling that answer questions like: By how much will decisions improve when additional information becomes available to reduce selected uncertainties? Which uncertainties should be reduced first? What data should one collect?

What was the key experience that made you want to do research in data science and what fascinates you about working with MDSI?

I work in a field where informative data is scarce, even today, due to our focus on extreme events. For instance, a 100-year flood occurs on average only once in a century. No matter how much data we gather, information on such rare events will always be limited. This scarcity is even more pronounced in structural engineering, where the rate of catastrophic failures in buildings and infrastructure is very low. Therefore, it is essential that we make optimal use of the data that we have when predicting risk in such systems. Standard statistical methods are typically insufficient, which early on in my career motivated me to investigate specific methods such as Bayesian networks. I hope that the MDSI will expose me to novel methods that can be employed in our context.

What paradigm shift do you expect within the next ten years, triggered in particular by the Institute's interdisciplinary approach?

In my field, predictions have traditionally relied on mechanistic models with randomized parameters to account for uncertainty. When relevant data is available, it is integrated into the analysis through Bayesian updating. Thanks to increased data-collection capabilities and advances in machine learning, novel ways of combining engineering and physical models with data in a probabilistic framework have emerged. Nevertheless, for the foreseeable future my field will continue to rely on a Bayesian philosophy, which acknowledges the subjective nature of the probabilistic prediction. This setting and approach differs from mainstream data science and machine learning approaches.  The MDSI can help us in motivating colleagues from computer science and mathematics to address this challenge.