Deep Learning For Scientific Discovery Group

Ramsés J. Sánchez is a theoretical physicist and applied statistician at the Lamarr Institute, where he leads the Deep Learning for Scientific Discovery (DL4SD) group and also serves as a postdoctoral researcher and scientific coordinator. He wrote his PhD dissertation on anomalous transport and out-of-equilibrium dynamics and received his doctoral degree from the University of Bonn. His research focuses on foundation models for zero-shot inference of stochastic processes and on phenomenological models for formal and commonsense reasoning with large language models.

David Berghaus is a postdoctoral fellow at the Lamarr Institute. He wrote his PhD thesis on the numerical computation of modular forms on noncongruence subgroups. His current research focuses on natural language processing and machine learning for science in both academic and industrial contexts. For additional information, please visit: https://david-berghaus.github.io/

Kostadin Cvejoski is a Machine Learning Researcher at JetBrains. He completed both his master’s and PhD at the University of Bonn. His doctoral research focused on addressing temporal distribution shifts in large language models, by incorporating temporal information. Currently, his research interests include: Representation learning for reasoning within foundation language models; Development of foundational models for time series data; Integrating temporal context into large language models; and Applications of machine learning and deep learning in human brain interaction.

Patrick Seifner studied economics and mathematics at the University of Bonn, focusing, respectively, on game theory and abstract algebra. His master’s thesis investigates some cellular structures of Hecke algebras and their connection to its representation theory. Since 2021, he works as a PhD student in machine learning (ML) at the University of Bonn, and member of the Hybrid ML group of the Lamarr institute. In his research, he explores deep learning methods for the inference of continuous-time stochastic processes from (potentially noisy) data.

Lea Busse is currently pursuing a Bachelor’s degree in Physics at Ruhr University Bochum. She has a keen interest in particle physics and the application of neural networks. Her passion lies in interdisciplinary research at the intersection of physics and computer science, where she aims to leverage computational techniques to advance the frontiers of physics.

Luca Eichler studied Computer Science at the University of Bonn. He conducted his master’s thesis in the DL4SD group, where his research focuses on amortized machine learning methods under model misspecification.

Manuel Hinz is a PhD student in Mathematics at the University of Bonn (INS, UKB), specializing in mathematical modelling and mathematical biology. He earned his Bachelor’s and Master’s degree in Mathematics at the same institution, focusing on analysis, numerics and probability theory, statistics, respectively. His research interests include high-dimensional stochastic processes, particularly their connections to statistics and machine learning. He is particularly interested in problems related to model fitting and sampling.

He is leading the development of the group website, especially the tutorials and worked on his masterthesis about the inference of aggregated Markov processes in the patch clamp view of ion channel gating during his time at Lamarr.

You can find his website here.