Ben Cox

Ben Cox

Computational statistician · Max Planck Institute for Physics

Pronouns: they/them

I am a computational statistician. My background is in spatio-temporal Bayesian signal processing, particle filters in particular, and the awkward business of estimating their parameters.

I did my PhD in statistics at the University of Edinburgh, and I am now a postdoctoral researcher at the Max Planck Institute for Physics, where I am loosely associated with the MADMAX and LEGEND groups. My current work is on model serialisation: developing software layers for the creation, transmission, evaluation, and long-term storage of statistical models.

This site holds my writing (rambling), my (incomplete) CV, and (a subset of) the projects I maintain.

Ask me about

Time series modelling, probabilistic programming, particle filters, Discworld, Evangelion (the original, not the rebuilds), my on-again-off-again (currently off-again) relationship with JAX, flight simulators, or anything that you are willing to talk about for hours while I take notes.

Languages

I primarily speak Python and Julia, though some C, C++, and Rust have managed to worm through my defences. I used to speak R, MATLAB, and Fortran, less so these days.

Human languages: native English, and German very much in progress. I can follow rather more than I can produce, and I would not claim more than that. Russian was somewhere around B1 about a decade ago, and has degenerated since.

In German, sie if you are feeling up to it, or just use my name. I am not in the habit of correcting people about this though.

Outside work

Ballet, Russian method, which I am currently relearning.

Custom hardware, mostly things I want to exist and cannot buy:

Riflery. If you know a good club near Munich, do get in touch.

Mountaineering, which I need to get back into shape for.

Tabletop roleplaying games, where the language barrier remains an obstacle.

Ongoing side quests: Deutsch lernen, physics, and whatever machine learning projects I cook up in my spare time, trees for preference. Programming for GPUs, largely so I can tailor methods to that paradigm: branchless, fixed working memory.

Contact

Funding

My position is funded by Germany’s Federal Ministry of Research, Technology and Space (BMFTR) within the ErUM-Data programme, under grant FKZ 05D25PC1 (DEMOS consortium).