Hi there 👋

I’m a postdoctoral researcher at Origins Data Science Laboratory (ODSL) in Munich and the Munich Center for Machine Learning (MCML). I did my PhD in in the Observatory of LMU and the Munich Center for Machine Learning.

Generating matter density fields with baryons conditioned on dark matter density fields with flow matching.

The diffusion process showing both the stochastic and deterministic paths through the marginal distributions of the diffusion process for a set of datapoints.

What do I do?

I like using generative models in Bayesian inference problems to extract information on fundamental physics in cosmology.

Right now i’m working on

• a simulation-based inference (SBI) package in jax,
• testing the limits of simulation-based inference with the one-point matter PDF,
• baryonification using generative models with physically motivated latent spaces,
• field-level inference pipelines using generative models.

I’m interested in generative models…

• Score-based diffusion models,
• Transformer Flows,
• Flow Matching,
• Variational Diffusion models (VDMs),
• Variational Diffusion Autoencoders (VDAEs),

…transformer models & geometric deep learning…

• Set Transformers,
• Compact Convolutional Transformers,
• Vision transformers,

…and statistical problems in general…

• Frequentist-matching priors,
• Information-maximising neural networks,
• Extreme data compression while searching for new physics,
• Hierarchical Bayesian Neural Networks (HBNNs).

I also teach MSc Physics students in the Physik x AI labs at LMU Physik where I write teaching material that delivers machine learning insights from problems in physics.

My goal is to show students cutting edge algorithms and statistical methods that they will not learn anywhere else.

• Simulation-based inference with random fields in cosmology,
• Generative models for Bayesian inference in cosmology (link not available).