Mateusz Kapusta (@Wesenheit)

Physics & ML research @ Warsaw University

Hi there!

I am a physicist and a software engineer with a passion for applying Machine Learning, Bayesian modeling, and High-Performance Computing to natural sciences. This webpage contains various content that I consider interesting enough to share with the world. You can find here my publications and blog posts.

Featured Content

--> IrisML - Neural Posterior Estimation for Spectral Energy Distribution fitting (2025)
  • Authors: Mateusz Kapusta
  • Journal: ICML 2025: Machine Learning for Astrophysics workshop Link

Fast and reliable real-time inference for bayesian SED fitting? Now possible with Simulation-Based Inference. Variable-size inputs supported! (Poster link)

--> GMVAE clustering applied to RNA sequencing (2023)

Traditional clustering methods typically rely on the raw, observed features of data. However, we can often achieve better results by using Variational Autoencoders (VAEs) to create an embedded representation of the data first.