My background is in theoretical physics (MSc) and computational neuroscience (PhD). I was fortunate to work with Klaus Hornberger on the theory of interference of entangled material particles for my MSc and for my PhD with Christian Leibold on models of memory sequences, combining theoretical neural-network models and data analysis on in-vitro recordings. In between I worked at the German Aerospace Agency and the European Southern Observatory.

During my first postdoc in Oxford I attended Nando de Freitas' and Phil Blunsom's lecture and grew interested in machine learning, complementing the classical signal analysis techniques with new Bayesian time-series models in work with Diego Vidaurre. After Oxford, I spent a few years between academia and industry, figuring out the right amount of theory and application, and which field in Science I liked most. It turns out I liked most of them!

Mach-Zehnder interferometry on entangled material particles.

Mach-Zehnder interferometry on entangled material particles.

Mean-field models of sequence replay with inhibition.

Mean-field models of sequence replay with inhibition.

Unsupervised segmentation on neural time series with a Bayesian hidden-Markov multiple autoregressive model.

Unsupervised segmentation on neural time series with a Bayesian hidden-Markov multiple autoregressive model.

Time-domain peeling reconstruction of complex synaptic events.

Time-domain peeling reconstruction of complex synaptic events.

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/30fa65d8-acfc-4ffd-acae-d33edd96f017/Untitled.png

My last station was with Jakob Macke, where I could work on what, for the physicist in me, is a very happy synthesis: unleashing the power of machine learning on mechanistic models in order to solve the inverse problem. So-called simulation-based inference methods aim at a Bayesian characterization of the parameters of a mechanistic model that make it reproduce actual observations.

https://www.mackelab.org/sbi/static/infer_demo.gif

As leader of the ml ⇌ science colab my vision is to realize a fruitful, fun, and intellectually rewarding exchange with with methodological and domain scientists to translate progress in machine learning into scientific discoveries.

Stations

2008-2012

PhD in computational neuroscience (Christian Leibold, LMU Munich)

2011-2015

MRC Fellow in the neural basis of memory (David Dupret, MRC and University of Oxford)

2015-2016

LMU Research Fellowship on statistical models of spatial firing (Andreas Herz, LMU Munich)

Visiting researcher, Human-Environment Relations in Urban Systems (Claudia Binder, EPFL).

2017-2020 Work at tech startups and as freelancer.

2018-2020 Research associate on simulation-based inference with Jakob Macke

Publications

See Google Scholar.

Free time

Football, cycling everywhere, back-country skiing and kiteboarding.