About me
After earning my Professional Bachelor degree in chemistry from Arthesis Plantijn Hogeschool, I continued to pursue a Master's degree in chemistry at the Vrije Universiteit Brussel. While working towards this goal, I discovered the world of computational chemistry and cheminformatics. Together with my interest in computers, it was only a small step towards artificial intelligence. I did my Master's dissertation about predicting reactivity indices from conceptual density functional theory with neural networks and developed AI methods for Multi-parameter optimization of formulation problems during my internship.
In the past I held a teaching position at Don Bosco Hoboken. Currently, I work as a Research Data Engineer at VITO in the SPOT team. Here, I focus on applying AI techniques in chemistry-related problems.
Predicting Reactivity indices
For my Master's dissertation, I worked towards developing convolutional neural networks for the prediction of conceptual density functional theory reactivity indices like chemical potential and chemical hardness. Models were obtained with R2 values up to 0.92!
Multi-parameter optimization
While working as an intern at 3M, I developed a methodology to optimize monomer compositions for synthetic polymers in function of certain properties, the glass-transition temperature among which. This method allowed for multi-parameter optimization and resulted in more accurate optima while providing significant speedups in research time. This methodology is applicable for both mixture design and full independent variable analysis.
Publications
- Jacobs, M.; Vermeersch, L.; De Vleeschouwer, F. Conceptual DFT Meets Machine Learning: A New Route to Enhanced Diels–Alder Reactivity. Journal of Computational Chemistry 2025. https://doi.org/10.1002/jcc.70277
- Desmedt, E.; Jacobs, M.; Alonso, M.; De Vleeschouwer, F. Deciphering Nonlinear Optical Properties in Functionalized Hexaphyrins via Explainable Machine Learning. Physical Chemistry Chemical Physics 2024. https://doi.org/10.1039/d4cp03303e.
Software: Polymer optimization
In order to further improve the proposed methodology for multi-parameter optimization and allow other colleagues to benefit from it, a small application with an interface was built. This standalone application could be used to design and optimize not only mixtures but also independent variables and incorporates the developed machine learning methodologies.
Software: Alchemical DFT
Upon completion of my Master's dissertation, the obtained models were bundeld in a small web application to be available for the larger public.
Software: CSMS: The Chemical Space Management System
CSMS or the Chemical Space Management System was designed and built to provide easy management of molecules and their quantum chemical data. It allows to build datasets and provide real time insight into the chemical space of that dataset. Afterwards, this data can be easily aggragated other to systems for training models. It is currenty in the development phase but a demo is available on request.
Software: Budo Management
The YNBK is a budo-related organisation that focuses on iai-jitsu, or the art of drawing the sword. I already train this as a hobby for a few years and hopefully many years to come. For this organisation, I wrote a web application to mange the members and allow instructors of other dojos to manage and grade their members. Check it out here.
