Hi there!
I am now a research associate in geometric generative models applied to molecular data, working with Michael Bronstein in Oxford.
Before returning to academia, I used to work on uncertainty quantification and free energy perturbation with my beloved MLR team at Bayer in Berlin.
I did a short postdoc at Aalto University with Vikas Garg and Markus Heinonen. I was incredibly lucky to do my PhD with Søren Hauberg and Georgios Arvanitidis, on differential geometry in machine learning. You can read my thesis here.
I tend to be a serial-workshop organizer with GRaM 2024, TAG-DS 2025, GRaM 2026, NeurReps 2026.
Oh, and I am a huge information-geometry nerd. I fell down that rabbit hole after reading Frank Nielsen’s work back in 2018 and never really climbed back out. If you somehow ended up here, and feel like nerding out, please say hi! I am based in Oxford, and sometimes in London, Berlin, Nantes, Copenhagen, Amsterdam and Vienna.
Bluesky / Twitter / Google Scholar / LinkedIn / Email
I am sad I left my beloved MLR team at Bayer. But the geometric calling was way too strong, and I am looking forward to working on geometric generative models with Michael Bronstein, in collaboration with Oxford and AITHYRA. If you are excited too, have a look at the research of Iolo, Manuel, Jacob and Teodora. ♥
Federica led a brilliant group of students working on evaluation metrics on Riemannian manifolds. Their workshop paper got accepted at GDDL at NeurIPS in Paris. So much love to them, and so happy I could help too. :)
NeurReps received more than 350 submissions across the extended abstracts, proceedings and findings tracks (!)
Released minimol-onnx. After struggling with too many environment incompatibilities, and inspired by cddd-onnx, I decided to make my own minimol-onnx. ONNX here should really be the gold standard when deploying a model.
NeurReps will return to NeurIPS (Sydney). More information coming soon.
The GRaM 2026 proceedings are on their way to publication in PMLR (volume 326). It was also a joy to gather the results of the GRaM competition into a meta-analysis paper. :)
GRaM 2026 at ICLR — our second edition! Proceedings coming soon.
On the Fisher Geometry of Diffusion Models’ Latent Space accepted as a poster at ICLR GRaM 2026. Led by Maria, a lovely follow-up to Rafał’s spacetime paper.
ICLR accepted two papers: The Spacetime of Diffusion Models: An Information Geometry Perspective (Oral) and Riemannian Variational Flow Matching for Material and Protein Design (Poster). A pleasure to work with the brilliant Rafał, Olga and Floor ♥
GRaM got accepted as a workshop at ICLR, our second edition! We have a call for proceedings, tiny papers, blogposts and a competition coming soon.
Gave a talk at GPSS on Links between Riemannian geometry and Gaussian Processes. Thank you Carl Henrik Ek and Mauricio Alvarez for the invitation!
Started at Bayer in the MLR team, working on uncertainty quantification, BO and AL for small molecules and FEP prediction.
End of the wonderful workshop on Geometry for statistics and AI, in the beautiful Metochi Study Center. Thank you Libby Baker, Stefan Sommer and Erlend Grong, for the invitations. We had so many exciting discussions and… wine. :)
Gave a talk at the “Geometry and Information” seminar at HSE, organized by Maxim Beketov. Loved the Saint Petersburg Mathematicians and Their Discoveries book gift.
First run of our Geometric Deep Learning course at Aalto! Created together with Caglar Hizli, Katsiaryna Haitsiukevich and Erik Schultheis. So many lovely and smart students!
Started as a postdoc at Aalto University with Vikas Garg, Samuel Kaski and Markus Heinonen. Excited for a new chapter!
PhD defence day — extremely nervous, but so lucky to have the most amazing examiners: Frank Nielsen, Michael Bronstein and Mikkel N. Schmidt. And the most incredibly supportive mentors and PI ever: Søren Hauberg, Georgios Arvanitidis and Carl Henrik Ek. ♥
Organised the Hitchhiker’s Guide to Complex Geometry reading group with the wonderful Pablo A. Morales (Araya Research) and Challenger Mishra (Cambridge University).
Started a 3-month internship at Relation Therapeutics. Together with Thomas, Alice, and Paul, we built PyRelationAL, a Python active learning library.
Co-organised CogSys ML Talks at DTU with Federico Bergamin: a seminar series on machine learning and applied mathematics. Lovely community building during covid, and some speakers ended up working together. :)
Start of my academic journey and my PhD… with the amazing Søren Hauberg at DTU. ♥