I am a postdoc at Delft University of Technology, the Netherlands. My research interests include scientific computing and deep learning with a focus on generative models, uncertainty quantification, data assimilation, physics-consistent machine learning, and fluid dynamics.
I am working as a postdoc in UrbanAIR project. The goal of the project is to develop a framework for urban air quality and heat dynamics forecasting. My role is specifically focused on the development of data assimilation and uncertainty quantification methods using deep learning techniques.
Together with Benjamin Sanderse, I developed generative models for physics applicatiions. The aim was to perform probabilistic forecasting and posterior sampling that adhere to the underlying laws of physics.
I am the lead developer of the Spatialise AI platform, SOCMO. My role includes the development of MLOps pipelines. This includes data science aspects such as model training, testing, and hyperparameter tuning. Furthermore, I am responsible for the ML engineer aspects such as model deployment, monitoring, and scaling.
My PhD project dealt with deep learning for data assimilation and inverse problems in physics applications. The aim was to perform real-time data assimilation with uncertainty quantification using deep learning techniques. My supervision were Cornelis Oosterlee and Sander Bohté.
The research project dealt with low noise supercontinuum sources for ultra-high resolution 800nm optical coherence tomography for glaucoma diagnosis. I was working on GPU acceleration of a C++ implementation of the 4th order Runge-Kutta Interaction Picture method to solve the generalized nonlinear Schrodinger Equation as well as uncertainty quantification of the input sources.
My work was included programming and mathematical modelling of various elements within the wind energy sector. Examples are time series models for weather with the goal of predicting production time of a wind turbine farm and analyzing buckling capacity of soil supported structures using partial differential equations and optimization techniques.
Experimentarium is a science museum, mostly for children and young adults. My job consisted of developing and performing science shows and experiments in front of large crowds and make complicated phenomena understandable and comprehensible for the layman and school classes.
Published in Scientific Reports
Published in International Journal of Greenhouse Gas Control
Published in the special issue Scientific Machine Learning: Blending of traditional mechanistic modeling with machine learning methodologies in the journal Computers & Mathematics with Applications
Published in Sensors
Published in Proceedings of the 37th International Workshop on Water Waves and Floating Bodies
Published in Journal of Computational Science
Published at Conference on Decision and Control (CDC)
My research interests include numerical methods for partial differential equations, machine learning, and their applications in computational science and engineering. I am particularly interested in the development of generative and reduced order models for parameterized partial differential equations using deep learning techniques and applying this to efficient real-time solving of inverse problems and data assimilation with uncertainty quantification in physics applications.