My main study focus are first-order methods [Bec17] under stochasticity, (total) asynchronicity assumptions and their application to parsimonious models for low-complexity inference. In that, I'm interested in Lyapunov stability, invex functional structures and cardinality constraint relaxation for sparse learning (e.g., for tree structured filterbanks), operator sparsification and factorization.

Before falling into the opt stew, I felt the joy of safe abstraction in projects with the Rust language. I'm a supporter of the Rust Machine Learning Group, past maintainer of Linfa [SLL23] and currently working on safe automatic differentiation compiler support for rustc.

I'm also host and user of the zettel.haus Mastodon instance for connected thoughts and an avid Vim and Linux user.

Besides that I really like making music. I play clarinets and currently learning a bass viola da gamba for diversion. I'm also interested in afro-american vernacular dances such as Shag, Lindy Hop or Balboa.

I studied electrical engineering and computer science at RWTH Aachen university.

---

[Bec17] Amir Beck. First-Order Methods in Optimization. Society for Industrial and Applied Mathematics, Philadelphia, PA, 2017. [ bib | DOI | arXiv | https ]
[SLL23] Lorenz Schmidt, Yuhan Lin, Rémi Lafage, Luca Palmieri, Ivano Donadi, et al. Linfa: A rust machine learning framework. https://github.com/rust-ml/linfa, 2023. [ bib ]