I am a first year PhD student at Aalborg University (AAU), advised by Prof. Kamal Nasrollahi and Prof. Sergio Escalera. My research interests converge at the intersection of Hyperbolic Learning, Concept Representation, and Machine Unlearning. My recent contributions have focused on exploring the emerging paradigm of Hyperbolic Machine Unlearning for concept removal in contrastive learning frameworks.
In 2024, I completed dual Master's degrees: an M.Sc. in Mathematics and an M.Sc. in Data Science from University of Barcelona (UB). I developed my Master's theses within the HuPBA group, led by Prof. Sergio Escalera and jointly supervised by Prof. Carles Casacuberta. My research investigated the foundational principles of Variational Autoencoders and Diffusion Models, where I proposed a Binary VQ-VAE as an expressive method for encoding motion in latent space. The aim was to efficiently generate human motion from text descriptions following a Bernoulli Diffusion process. Results didn't fully meet expectations, but the project was tremendously fun. Since then, I've been exploring various approaches to create more expressive latent space embeddings.
