Problem Solver, Modeller, Programmer, AI Expert

Modern machine learning applications, including but not limited to neural networks for language processing, rely on pattern recognition in suitably large example data. In my work, I focus on the verification of successful AI training, and the active acquisition of supplementary training data if needed.

Combining various areas of physics is necessary to understand surface or interphase processes, but introduces unique problems for solving the resulting equations. During my time at the German Aerospace Centre, I learned the intricacies of combining fluid dynamics, mechanics, reactions, and thermodynamics in one simulation.

Predicting the weather, understanding transport bottlenecks, or operating batteries benefit from a thorough consideration of the effects hidden within incomplete sensor data. With a broad spectrum of methods from data assimilation over stochastic processes to model selection, I can model and steer digital twins for industrial processes, logistics, or spreads of infectious diseases.

At the first international battery engineering conference after the Covid lockdowns, I met Masaki Adachi, then PhD candidate at the University of Oxford, and (as of 2025) a research group leader at Toyota in Japan. His work on Bayesian Optimisation and especially Bayesian Model Selection revolutionised the way we set up and validate battery models for material characterisation and operation prediction. I contributed my experience in the practical application of related Bayesian approaches to battery modelling tasks, which he incorporated in his work to bridge his groundbreaking algorithm BASQ to practical application with SOBER. I wrote a clear interface and documentation for the usage of these two algorithms, and wrote their inclusion in the battery model characterisation and selection software PyBOP.

At the flagship Norwegian research institute SINTEF, the research group around Simon Clark works on what I would call an "international battery materials science glossary", the ontology BattINFO. It is a concerted effort within the broader EU battery research context to harmonise the experimental data from characterisation measurements. This is the first step towards utilising AI to find patterns in battery materials science, i.e., it is the basis to even formulate a training dataset for LLM approaches with scientific accuracy. My contribution is a proof-of-concept of how to sustainably generate research data that complies to the high standards of BattINFO and the FAIR principles in general. I formulated a complete and automatisable workflow from battery measurement data to battery materials characterisation using the research data management and database software Kadi4Mat, incorporating the expertise of various experts along the way.

My personal project exploring the requirements for battery materials characterisation and showcasing automation solutions for that. As part of two of my scientific publications and subject to a third one, it is not under active development anymore and serves as their supplementary information now. The materials characterisation aspect is superseded by PyBOP and SOBER, to which I contributed the lessons I learned with EP-BOLFI and its relevant components. The automation aspect still stands on its own, but is now part of the larger EU project DigiBatt, and work that contributes to automation efforts continues there.

With the framework of Bayesian optimisation, I complement the effectiveness of neural network solutions with quality control. While the flexibility of neural networks allows them to learn pretty much anything, hallucinations are challenging to predict and prevent ahead of time. This may be fine for chatbots, but not for safety-critical engineering problems. Thus, I apply the same learning paradigm from neural networks with a different methodology: gaussian processes, or in general, stochastic processes. These implement the field of stochastics, which is the mathematical study of randomness and uncertainty. I studied and contributed to their application, which is initially more computationally intensive, but rewards us with clear predictions for the extent of random effects.

During my PhD at the German Aerospace Centre, I presented my research at various international conferences. I learned how to present my work comprehensibly to a broad spectrum of audiences, and how to understand their equally broad spectrum of research: surface engineering, quantum chemistry, mechanical engineering, IT, AI, and fluid dynamics, to name a few.

As I worked with chemists, physicists, mathmaticians, mechanical engineers, computer scientists, and industry experts to solve current research questions in battery engineering, I learned how to work within, lead, and coordinate a team comprising various backgrounds.

Deploying any software solution that is user-friendly involves a stack of technologies that powers it. When I deploy collaboration platforms, electronic laboratory notebooks, or database systems in general, I always study the tech stack from top to bottom to be able to adapt it to user requirements.

• Python
• C++
• Git
• SQL
• Bash
• Docker
• PHP
• Javascript

Battery Materials Researcher

• modelling of multiphysics systems comprising diffusion, reaction, flow, or thermal effects
• optimal control and state estimation for lithium-ion batteries and fuel cells
• reduced-order and asymptotic modelling to compress physical equation systems for limited computational resources
• training colleagues in new software stacks
• interdisciplinary collaboration between mathmaticians, physicists, chemists, computer scientists, and philosophers
• PhD, formally in chemistry (pending)

M. Sc. Mathmatics (very good, 1.4), minor subject Physics (~ B. Sc.)

• simulation and modelling of physical systems
• stochastic methods for digital twins
• numerical solution methods for partial differential equations and

Abitur (very good, 1.4)

Otto-Hahn-Gymnasium Ludwigsburg

2004-2012