Computer Science · National University of Singapore Contact: ryoji[last_name][at]u.nus.edu
I’m a second-year PhD student at the School of Computing, National University of Singapore, advised by Xavier Bresson.
Previously, I worked as a research fellow at New York University Abu Dhabi, advised by Djellel Difallah. I received my Bachelor of Science from NYU Abu Dhabi, majoring in Computer Science and Mathematics.
My research interests are AI for science, graph machine learning, and deep learning. I’m especially interested in applying machine learning to materials science. Can we build models that respect physical constraints and generate materials that can be synthesized?
DBRAG was accepted to the AACL-IJCNLP main conference. See you in Hengqin!
RAW-Explainer was accepted to Learning on Graphs. See you in Boston!
Selected for the NUS School of Computing Teaching Fellowship Scheme.
Started my PhD in Computer Science at the National University of Singapore.
AACL-IJCNLP 2026 · Accepted, main conference
DBRAG answers questions whose evidence is spread across multiple database tables. It retrieves and reranks relevant tables, then uses program-aided reasoning over their full contents while keeping the language model’s initial context compact.
Learning on Graphs 2026 · Accepted
RAW-Explainer explains knowledge-graph link predictions by finding small, connected subgraphs through a learned random-walk objective. A separate robust evaluator helps assess these explanations when their distribution differs from the full graph.
Preprint · 2026
We learn reusable crystal concepts that capture local atomic environments and global symmetry, then recombine them to guide the generation of new materials. This gives more control over exploration beyond known structures and improves the discovery of novel crystals.
Preprint · 2026
Crys-JEPA learns crystal representations that preserve differences in formation energy, allowing candidate stability to be assessed through embedding comparisons. Promising candidates are then used to refine the generator, improving the search for crystals that are both stable and novel.
SIAM International Conference on Data Mining (SDM) · 2024
XGExplainer improves graph-neural-network explanations by training a separate evaluator on randomized subgraphs. This helps existing explanation methods recognize informative subgraphs even when they differ from the data used to train the original prediction model.
Selected as one of two Computer Science PhD students recognized for exceptional teaching.
A highly selective research fellowship for graduating NYU Abu Dhabi students.
Teaching Assistant · National University of Singapore
Teaching Assistant · National University of Singapore
Teaching Assistant · National University of Singapore
Teaching Assistant · NYU Abu Dhabi
National University of Singapore
Committee member · NYU Abu Dhabi