I defended my PhD at the TU Berlin Machine Learning Group under Klaus-Robert Müller and Manfred Opper in November 2024 and got a nice new hat 🎓.

The title of my thesis was ‘Time-Reversible Bridges of Data with Machine Learning’. In my PhD I considered different state spaces and trained neural networks to recreate time-reversible dynamics in these state spaces with boundary conditions. The dynamics in my PhD included continuous and deterministic, discrete and stochastic, and finally continuous and stochastic. The applications were modelling MD simulations and quantum chemical Hamiltonians with recurrent neural networks, discrete state space diffusion with birth-death processes, and solving the Schrödinger Bridge Problem with neural networks. A lot of the topics I spent time on can be found in generative models, e.g. velocities fields in flow matching (deterministic and continuous dynamics), SDE’s in diffusion models (stochastic and continuous dynamics in Schrödinger Bridges) and Continuous Time Markov Chains (stochastic and discrete half bridges).

Over the course of my studies I gravitated towards machine learning fields with rigorous mathematics. Before I knew it, I ended up working on stochastic processes and stochastic differential equations. In particular, I have trained neural networks in stochastic differential equations in order to solve optimal transport problems and infer the corresponding stochastic process. The stochastic processes I tackled include SDE’s, CTMC’s, the connection between SDE’s and CTMC’s in generative modelling and the Schrödinger Bridge Problem. I especially enjoy the mathematical rigor of stochastic processes and probabilistic machine learning in general.

Currently I’m spending time at Prescient Design within Genentech/Roche designing and implementing advanced sampling and sample optimization schemes for protein generation and manipulation with machine learning models.

Besides my research activity, history in conjunction with economics and politics has been a long standing interest of mine. As I grew older the link between economics and history became more evident to me and my focus shifted to economic and financial history which often does a superb job at explaining the latent forces that shape history.

I have been remarkably lucky to be living in Berlin where I explore the multitude of different subcultures that exist so harmoniously in this city. Berlin represents something of an ideal city for me as its history has enabled subcultures to thrive while offering abundant intellectual goodies ranging from art&music, a lot of techno, to blockchain and machine learning.