Towards a Better Handling of Uncertainty

You will probably get various answers to this question if you ask different people. Some will equate it with machine learning, but for me data science encapsulates many methods that are essentially designed to make use of large data sets in a useful way. Data Science is fundamentally interdisciplinary, bringing together scholars and approaches from different fields like mathematics, statistics, engineering and computational sciences with a focus on real-world-scenarios. My personal research interests primarily lie in better understanding how new data science and machine learning based methods work, and in developing practically useful yet mathematically rigorous theorems that shed light on the inner workings of these methods.

Through my career I have had an interdisciplinary background. I did my PhD in environmental engineering, where I was working on data assimilation algorithms to provide better hydrologic and meteorological forecasts, e.g. with respect to flood events. I began to become more and more interested in the mathematical and statistical side of these methods, which then led me to undertake a postdoc in mathematics, where my research focused on understanding the theoretical properties of a new class of data assimilation methods. At times I felt a bit identity-less due to my interdisciplinary background, because in Academia there is a strong tendency to put people into specific boxes,. I found it hard to put myself firmly into one box, and luckily Data Science has given a home to my various interests. With my postdoc training, I’m trying to combine the different aspects of my career, which gets me engaged into statistics, analysis, machine learning as well as specific applications.

Regarding Hydrology, I have a strong background through my Bachelor’s and PhD studies in environmental engineering with a concurrent degree in mathematics. I am well connected with the hydrology community. In the biomedical field, I’ve been involved in a project related to blood flow modeling through the brain, aiming to predict aneurysm formation.  A particularly challenging aspect of blood flow modelling is uncertainty quantification.  Together with my collaborators – computer scientists and CFD modellers – we have developed a new method to improve uncertainty quantification in one of the most challenging parts of cerebral blood flow modelling: model geometry specification.  This collaboration was particularly fruitful as our differing expertise helped to address both the theoretical, computational and medical application aspects of this project.

The YAS gives me the chance to interact with scholars from fields of research that I have not interacted with before: from the Humanities, Politics or Anthropology for example. And although I am situated in Australia and other members across the globe, we have a vital exchange. Together with my YAS colleague and Historian Javier Francisco, we set up a project on understanding the historical evolution of environmental degradation and slavery under colonization in the Caribbean. Traditionally, historical research in this area has involved studying individual case studies for different Caribbean islands, relying on historical sources like documents and maps. Our goal is to identify common patterns across multiple islands in the Caribbean. Using methods from the Data Sciences, we seek to construct a model that can explain these patterns on a broader scale, encompassing the entire Caribbean region instead of isolated case studies. For me, those experiences from different fields of application are very fruitful because it helps in motivating my theoretical work. And of course, it broadens my personal horizon.