Probabilistic modelling
Bayesian inference, Gaussian processes, active learning, uncertainty quantification, and simulation-based reasoning.
Hi, I’m Jonas!
I make slow-to-compute problems fast.
I’m a quantitative researcher and scientific software developer with a background in cosmology. I build probabilistic machine-learning methods and reliable computational tools for Bayesian inference and gravitational-wave astronomy. I’m currently a postdoc at KBFI/NICPB in Tallinn.
From mathematical model to usable software.
My work usually starts with an expensive or statistically difficult problem and ends with a method that can be tested, documented, and reused. The same combination is relevant well beyond cosmology: uncertain data, costly models, and decisions made under a limited computational budget. I am also interested in how neural networks learn internal representations, what those representations encode, and how understanding them can make AI systems more interpretable and reliable.
Computational methods
JAX and Python implementations, differentiable models, accelerator-aware numerical work, and reproducible analysis pipelines.
Research software
Open-source packages with documented interfaces, validation, releases, and interactive explanations of the underlying methods.
Most of the time I’m a pretty chill dude. Outside work, I’m usually hiking, climbing, skiing, surfing, skydiving, or finding another questionable way to spend time in the mountains. I also make jewellery, which I share at @jeg_jewellery.