Dr Johannes Heyl

Dr Johannes Heyl has obtained a MSci in Physics with Theoretical Physics at Imperial College London and a PhD in 2023 in Data Intensive Science at University College London (UCL). His interdisciplinary PhD project was at the intersect of Astronomy, Chemistry and Computer Sciences. His work revolved around developing novel statistical as well as machine learning methods to better understand astrochemical processes. These processes are often underpinned by coupled systems of ordinary differential equations making the relationship between the inputs and outputs non-linear and difficult to understand. Cutting-edge machine learning interpretability techniques were able to provide interpretations to the relationships between physical parameters. Dr Heyl is now postdoctoral research associate at UCL.

Dr Heyl embarked on a PhD thesis project aimed to link astrochemistry and statistical and machine learning techniques, a completely novel approach for astrochemistry that traditionally had stayed away from these techniques. During his PhD, he demonstrated high levels of curiosity, interests and independence that allowed him to explore different new techniques or methods to aid astrochemical studies. His skills allowed him to publish 6 papers including one in a field completely different from astronomy (Data Science in Health), a remarkable achievement considering the requirement of a 6-month industry secondment in addition of taught courses.

Each article led by Dr Heyl has already had a high impact in the field. For example, his work on Bayesian Inference of reaction rate parameters as well as on the study of network topology. When modelling and predicting molecular abundances in the dense gas of the interstellar medium, one of the biggest challenges is the completeness and accuracy of the used chemical networks. The combination of techniques has opened up a completely new avenue for sensitivity analyses as well as reduction networks. He also laid out work on interpretable machine learning and showed a very novel and quick way to perform sensitivity analyses, allowing a real potential and rigour that traditional sensitivity analyses methodologies do not have. This was the first time that the concept of machine learning interpretability has been adopted in astrochemistry. The PhD thesis of Dr Johannes Heyl was conducted at at the Department of Astrophysics and the Centre for Doctoral Training in Data-Intensive Science at University College London, under the supervision of Professor Serena Viti.