PhD student at LIDIA · Universidade da Coruña
I work on machine learning, focusing on how AI can understand visual scenes and predict what happens next.
Personal website ↗ · Email · LinkedIn · X
PhD student at LIDIA · Universidade da Coruña
I work on machine learning, focusing on how AI can understand visual scenes and predict what happens next.
Personal website ↗ · Email · LinkedIn · X
Official PyTorch implementation of "Space-Aware World Models: Spatial Persistence Through Factored Scene Representations" (NeurIPS 2026).
Python
Modern PyTorch implementation of World Models with interactive notebooks for the Car Racing environment. Features VAE vision model, MDN-RNN memory system, and CMA-ES controller with visualization t…
Research code for my Final Degree Project (TFG) on perceptual latent representations for World Models, improved visual fidelity, latent dynamics, and autonomous control.
Jupyter Notebook 2
Residual-variance parent sampling on top of 3DGS-MCMC (gsplat 1.5.3). Benchmark on Mip-NeRF 360, Tanks & Temples and Deep Blending at 50K-200K Gaussians. Browser viewer with trained 50K-Gaussian mo…
Python
How much MLP depth does a Fourier-feature input encoding replace? A controlled study across 1D signals, images, 3D shapes and tabular data
Python
Forecast future weather conditions by comparing multiple deep learning architectures (DNN, CNN+DNN, LSTM, GRU, and Attention LSTM) on the Jena Climate Dataset. This interactive notebook delivers da…
Jupyter Notebook 3