At the University of Murcia, I work across the machine-learning research lifecycle, from data pipelines and model design to evaluation and open-source delivery. My work spans LLM fine-tuning, heterogeneous graph neural networks, and reliable machine learning.
I built Phenolinker, an explainable heterogeneous GraphSAGE system for phenotype-gene link prediction, combining BioBERT node features with Integrated Gradients. Its predictions contributed to real-world genetic diagnoses.
For GO3, I engineered a Rust core with a Python API, PyO3, and Rayon parallelism. Benchmarks showed faster initialization and gene-level similarity than established Python/R alternatives, while validating numerical agreement.
During a visiting research stay at the German Research Center for Artificial Intelligence (DFKI), jointly with Osnabrück University, I extended my doctoral work on representation learning across language, graphs, and structured knowledge. I’ve published six peer-reviewed papers, four as first author, and have also taught practical sessions in machine learning and deep learning.
