Celia Rubio‑Madrigal


About me

I am a PhD candidate at the Relational ML Lab, supervised by Dr. Rebekka Burkholz at CISPA Helmholtz Center for Information Security in Saarland, Germany.

I study ways to improve the generalization of graph neural networks (GNNs), with a focus on how graph structure and depth affect their behavior in theory and practice. I have worked on directions such as rewiring, tabularization, and knowledge distillation, with an emphasis on empirical evaluation and testing the limits of common assumptions. More broadly, I am interested in developing architectures that balance the trade-offs between expressiveness, efficiency, and generalization; in this regard, I have collaborated on broader topics such as tabular robustness and optimization. My work has been published at NeurIPS, ICLR, and ICML [1, 2, 3] and presented in several talks and workshops.

Background

I hold degrees in Mathematics and Computer Science from Universidad Complutense de Madrid. In 2022, I was awarded a postgraduate fellowship from la Caixa Foundation, which fully supported my master’s studies at the University of Strathclyde in the UK, where I was awarded the Departmental Best Overall Performance Prize.

I was featured on the Nova 111 Student List 2023 as one of the top 10 under 25 in Computer Science in Spain. I have also been selected twice for the Heidelberg Laureate Forum: the 13th edition in 2026 and the 9th edition in 2022, where I was interviewed.

Beyond research, I annually attend the European Girls’ Mathematical Olympiad (EGMO) as part of the IT team with Joseph Myers since 2023. In 2019, I developed a LaTeX/tikz package for twelve-tone music notation called ddphonism.

My CV is available here.

Upcoming

Recent

Older events are shown in the timeline.