Welcome to my GitHub profile! I'm a dedicated mathematician and AI engineer, learning a bit more everyday about Artificial Intelligence.
Some of the key contributions that cover the whole spectrum of what I do are:
- Contributed to the open source Machine Learning project Scikit-FDA by implementing, among others, a non-linear dimensionality reduction method for functional data. I created this website for an overview on how the method works.
- Lead version 2 of the project subwiz, a lightweight GPT-based transformer model designed to discover and enumerate subdomains.
- Worked on other programming projects, like Ad Astra C videogame and Crypto mining simulator, with friends.
The latest publications in which I participated are available in Link to Scholar. Here are a few (as of February 2026):
- Researched the intersection of neural networks and finite element methods in the following paper Can Neural Networks learn Finite Elements?.
- Develop a novel symmetry-aware autoencoder framework for neural network canonicalization that leverages Scale Graph Metanetworks (ScaleGMNs) to achieve superior model merging by jointly accounting for permutation and scaling symmetries. Symmetry-Aware Graph Metanetwork Autoencoders: Model Merging through Parameter Canonicalization
- In the field of zero shot composition for Reinforcement Learning (at the University of Amsterdam).
- In the field of Riemannian Geometry and model merging (at the University of Amsterdam).
- In the field of Query Answering in Knowledge Graphs (at the Vrije Universiteit Amsterdam).