Master's defense
Felipe Bevilaqua defends a master's dissertation on fairness in synthetic data
Synthetic tabular data amplifies classifier unfairness — yet standard mitigation still works when the generated data is high-quality.
Read the paper → COPPE / UFRJ · PESC · Rio de Janeiro
AI research worthy of trust.
SABIÁ is a research lab at PESC/COPPE-UFRJ. We investigate how learning models can be robust, fair, and auditable — and we train Brazilian researchers to build this new generation of intelligent systems.
We research
Trustworthy AI Robustness Recommender Systems Fairness Intelligent Systems Language Models
Research areas
How our fronts talk to each other.
Six main areas, dozens of sub-topics. Hover to see the specific cuts of each front — and where they intersect.
Research area Sub-topic Cross-area collaboration
↑ hover / tap a research area
Where we publish
Open science, at the venues that define the field.
NeurIPS
Machine Learning ICML
Machine Learning KDD
Data Mining WSDM
Web Search RecSys
Recommender Systems ACL
NLP + Workshops at FAccT, BRACIS and ENIAC · Open preprints on arXiv
People
Researchers training researchers.
General coordinator · Full professor Prof. Geraldo Zimbrão
Recommender Systems Intelligent Systems
Post-doc · Associate researcher Prof. Leandro Alvim
Fairness Language Models
Post-doc · Associate researcher Prof. Filipe Braida
Intelligent Systems Recommender Systems Robustness Language Models
Post-doc · Associate researcher Prof. Ygor Canalli
Trustworthy AI Fairness
Master's Rodrigo Pereira Pagliusi
Fairness
Master's Felipe Bevilaqua Foldes Guimarães
Fairness
MSc 2017 · PhD 2024 Julio César Barbieri Gonzalez de Almeida
Recommender Systems
Master's Felipe Bezerra de Melo
Fairness
PhD candidate Adriano Beringuy
Language Models Fairness
Master's student Anderson Soares Rezende
Language Models
Master's student Felipe Assis de Souza
Language Models
Master's student Julia Oliveira Melhorim
Intelligent Systems
Master's student Kevin Torres Ribeiro
Fairness
Master's student Leonardo Emerson André Alves
Language Models
Master's student Maria Bianca Monteiro Irace
Language Models Trustworthy AI
Master's student Maria Luiza Cantanhede Wuillaume
Master's student Pedro Gabriel Nunes Gadelha
Intelligent Systems
About
Frontier research, anchored in Brazil.
The SABIÁ Lab is a research group within the Systems Engineering and Computer Science Program (PESC/COPPE) at UFRJ, founded on one conviction: the next generation of AI systems must be built with the same rigor we use for critical infrastructure — bridges, power grids, financial systems.
We work across six fronts — from language models to recommender systems — always with the same underlying question: how does this system fail, and how do we know it is trustworthy? We operate in open science, publishing at the field's leading venues and partnering strategically with the public and private sectors.
- Program
- PESC · COPPE · UFRJ
- Location
- Rio de Janeiro, Brazil
- Lead
- Prof. Geraldo Zimbrão
- Venues
- NeurIPS · ICML · KDD · WSDM
News
What's in motion.
Master's defense
Rodrigo Pagliusi defends a master's dissertation on fairness
A reproducible stress-testing methodology to measure how classifiers degrade — in performance and fairness — as data bias grows.
Read more → PhD defense
Ygor Canalli defends a PhD on fairness in machine learning
Two new techniques for fair classification — Fair Transition Loss and Redlining Penalty Regularization — inspired by label-noise robustness.
Read more →