2027 → ?
Next state not observed yet. Target: a master's in robotics / machine learning, and research on world models for robots.

Maths & Physics → robot learning
Final-year student of the double degree in Mathematics and Physics at the University of Seville. I'm interested in robots that learn how the physical world works.
Now: Final year · research student (EDAN) · thesis on training dynamics · world models
















Open to robot-learning internships from summer 2027, and always happy to talk about world models.
Seville, Spain · EU citizen
Next state not observed yet. Target: a master's in robotics / machine learning, and research on world models for robots.

Since October 2026 I'm a research student in the Department of Differential Equations and Numerical Analysis at the University of Seville. The project, Invariant Learning Dynamics in Transformers, studies a theoretical framework for how inductive-reasoning abilities emerge in Transformer language models. The idea is to describe training through a low-dimensional invariant manifold instead of the full parameter space, so the problem can be studied both on paper and in experiments.

Building up from deep learning to world models: models that predict how the physical world evolves, so a robot can plan before it acts. I'm working through the foundations and moving into simulation (MuJoCo) and low-cost robot arms (LeRobot). This is where my next projects will appear.

My bachelor's thesis treats gradient-based training as a discrete dynamical system, where every optimisation step is one iteration of a map. Through that lens I study how choices like the learning rate change how a network trains, combining mathematical analysis with computational experiments on small networks in Python.

Overall band 8.0 (C1). My English started during a year living in Ireland at 12.

Certified B1 at the University of Münster after a year of lectures in German.

Certificate from the NVIDIA Deep Learning Institute on GPU-accelerated data science: processing and modelling data on the GPU instead of the CPU.

Selected for a tech-entrepreneurship programme led by Juan Martínez Barea. Bootcamp completed.

I spent 2025–26 at the University of Münster studying mathematics and physics. Lectures were in German and exams in English. Highlights were signal processing, experimental and theoretical physics, and advanced lab courses where I measured and analysed real data: atomic force microscopy, X-ray experiments and semiconductor characterisation.

I created Matteo, a recurring character made entirely with generative image and video models, and built a series around him. In a year it reached 400k followers on TikTok, 80k on Instagram and more than 80 million views.

I built one project each for six small businesses: two websites, two customer-service chatbots on the Claude API, and two review-management automations that run in the background. I handled each one end to end, from scoping with the client to building with Claude Code and n8n, connecting external APIs and handing it over.

A five-year double degree (360 ECTS). The parts I use most for robot learning: optimisation, numerical methods and simulation, differential equations, probability and statistics, and classical and analytical mechanics. That is the maths of how physical systems move and how models learn.
At 18 I moved to Zürich on my own for the summer and worked in construction (plastering, ceiling work) and removals for Keller bis Dach, Toggenburger, Pepa AG Bau and Immark AG. I went back the following summer. Hard physical work in a new language with no safety net: it's where I learned to figure things out on my own.

Scored 13.7 out of 14 in the university entrance exam (top 0.5%). I also finished school with Matrícula de Honor, which came with an Excellence Scholarship from the Junta de Andalucía.

Official French diploma, level B1.

Lived in Ireland for a year at 12. It's where my English really started.

Twelve seasons of federated football, from age 6 to 18. League champions in 2021, and captain in one of my youth teams.