I develop bio-inspired machine learning systems and physics-informed digital twins for engineering applications — designed to learn fast, generalize well, and run efficiently.
My work spans Liquid State Machines/Reservoir Computing (a biologically-inspired ML paradigm ideal for edge deployment and structural monitoring), optimal network design (how topology shapes learning), and nonlinear dynamics (bifurcation-aware modeling with minimal data).
At TU Berlin, I develop dynamics-informed learning systems for structural mechanics as part of DFG SPP 2353. Previously, I built a theoretical framework for biochemical information processing at the MPI, Bonn — work that now informs my approach to bio-inspired and neuromorphic ML architectures.
I’m interested in 🤝collaborations and roles at the intersection of efficient AI, nature-inspired neuromorphic computing, and physics-constrained learning.
The Reservoir Computing Conference 2026 (RCC26) is an international conference dedicated to the field of reservoir computing and related machine learning approaches. Co-organizing …
A Berlin University Alliance (BUA) Course for engaging students with hands on research by building a Physical ML device that run on water and helping young scientists lead research …
This project has been started to understand emergent structure-function relationship of evolving networks. Part-1: Performance-Dependent Network Evolution (PDNE) framework. I …
PyReCo is a Python based library built by researchers for researchers: we aim to develop new RC methods that allow for fast and efficient learning for sequential data. The main …
A Keynote Talk was presented in the Dynamics Driven Dynamics Session on the *Foundations, Theory and Applications of Reservoir Computing* at the 11th GACM conference in TU …
Performance-dependent network evolution is applied to Wilson-Cowan neuronal dynamics, revealing compact reservoirs that generalize well and recover interpretable …
Study of homeorhetic regulation mechanisms in cellular phenotype dynamics.
Can evolution beat random chances? Our answer: Yes, we have a proof! 😃
This study uses a data-driven approach to investigate how bifurcations can be learned from a few system response measurements.
We explore the effect of a common external system, which may be considered as a common environment, on the oscillation death(OD) states of a group of Stuart–Landau(SL) oscillators. …
Project Planning The next phase of of the project of information processing with bucket of water based Physcial Reservoir Computers is being planned to generate music with it. …
