NLP and graph representation learning
My current work at Cardiff studies NLP models that use graph representations and reasoning, with GPU-accelerated training and evaluation in PyTorch.
Postdoctoral Researcher · Cardiff University
John Stewart Fabila-Carrasco
Graphs, signals, and machine learning for complex data.
I study how structure can make complex data easier to model and interpret, from language and networks to biomedical signals. My work combines mathematical foundations with reproducible experiments in Python and PyTorch.
15 peer-reviewed papers
207 google scholar citations
23 research talks
250+ teaching contact hours
Research
Three connected directions
A common question runs through my work: how can relationships in data be used to build better models, measurements, and explanations?
Research overviewSpectral graph methods
I develop spectral tools for graph comparison and clustering, including magnetic Laplacians, spectral bracketing, and constrained graph clustering.
Graph signals and scientific data
I design entropy and complexity measures for signals on networks, with applications to EEG, fMRI, DTI, sensor, and flow data.
Applied data practice
Public-sector analytics and decision support
At INEE in Mexico, I worked with national assessment and census data, combining statistical modelling, GIS analysis, and dashboards for policy teams.
Experience
Research appointments
- Postdoctoral Researcher in NLP and Graph Representation LearningCardiff University
- Postdoctoral Researcher in Scalable Spectral ClusteringUniversity of Edinburgh, School of Informatics
- Postdoctoral Researcher in Graph Signal Processing for Biomedical DataUniversity of Edinburgh, School of Engineering
- Industry and Government Data Science ResearcherNational Institute of Educational Evaluation (INEE)
Training
Education
- PhD in Mathematical EngineeringUniversidad Carlos III de Madrid
- MSc in Mathematical EngineeringUniversidad Carlos III de Madrid
- MSc in Mathematical SciencesNational Autonomous University of Mexico
- BSc in MathematicsAutonomous Mexico State University
Selected publications
From spectral foundations to graph learning
Four papers that trace the main development of my research.
Complete publication list- 2025
Signed Laplacians for Constrained Graph Clustering
A graph clustering method that respects constraints while remaining mathematically principled and scalable enough to matter for real structured datasets.
- 2024
Isospectral Graphs via Spectral Bracketing
A theoretical paper on spectral graph structure that strengthens the mathematical foundations behind later graph-analysis work.
- 2024
Graph-based Permutation Patterns for the Analysis of Task-related fMRI Signals on DTI Networks in Mild Cognitive Impairment
A way to study how brain activity evolves across anatomical networks rather than treating signals as isolated time series.
- 2023
Dispersion Entropy for Graph Signals
An entropy measure designed for graph-shaped data, making it easier to quantify complexity in signals that depend on network structure.
Communication
23 research talks across nine countries
Invited and refereed talks in graph theory, signal processing, and network science.
Talks by yearTeaching
Mathematics, statistics, and engineering
250+ contact hours, three MSc co-supervisions, and a Teaching Excellence Award.
Teaching and supervisionContact
Research correspondence and collaboration
For questions about research, publications, or talks, email is the best way to reach me.