Rede BAh.IA
AI literacy for public education: an open curriculum with lesson plans, slides and guides for K-12 teachers, including schools in underserved neighborhoods.
+80municipalities reached
FundingSECTI · Bahia State Government
Institute of Computing · Federal University of Bahia
Frontier research in Artificial Intelligence.
LabIA is a research laboratory at the Institute of Computing of the Federal University of Bahia (UFBA). We develop methods in reinforcement learning, continual learning, graph-based learning and optimization, with applications in health, urban mobility and education.
- Reinforcement learning
- Continual learning
- Graph learning
- Optimization
CNPq Research Productivity Fellowsleading the laboratory
Group registered with CNPqDirectory of Research Groups
24 public repositoriesdatasets, environments and models
Projects
LabIA’s research underpins large-scale, publicly funded projects whose impact reaches beyond the university: teacher training, public education policy, innovation and health.
Instituto BAh.IA
An AI research, development and innovation institute selected under the INCITE program. It connects LabIA’s research with public and private partners and turns results into applied solutions.
7partner institutions
FundingFAPESB
I-CARE
Decision support for diagnosing hematological diseases such as multiple myeloma, with public bone-marrow image datasets published in Scientific Reports and Scientific Data.
+4.000image downloads
FundingMaria Emília Foundation
Lifelong Reinforcement Learning
Lifelong reinforcement learning for urban systems: agents that keep learning as the city changes, tested on digital twins of Salvador’s public transit.
Who funds LabIA’s research
CNPqFAPESBMaria Emília FoundationKunumi InstituteBahia State Government · SECTI
Research areas
Four methodological areas shape LabIA’s work. We develop theory and algorithms in each of them, and many of our most interesting results emerge where they intersect.
Reinforcement learning
Agents that learn to make decisions by interacting with their environment. We study cooperative multi-agent RL, Dec-POMDP formulations, partial observability and sparse rewards, always in environments that represent real systems.
Resources: BusEnv
Continual learning
Models that keep learning after deployment. We study catastrophic forgetting, concept drift and non-stationary data streams, combining machine learning with time series and dynamical systems.
Applications: urban mobility, clinical data
Graph-based learning
Graph neural networks for relational and spatio-temporal data. We model everything from interactions between protein residues to a metropolitan transit network, with attention to interpretability and scalability.
Optimization
Single- and multi-objective combinatorial optimization, evolutionary computation and hybrid metaheuristics. We also explore the frontier with learning: models that learn heuristics and optimization that improves model training.
Applications: resource allocation, routing, logistics
Where the areas meet
- RL on graphsMulti-agent policies that use the network topology as the communication structure between agents.
- Continual learning in dynamic environmentsAgents and models that adapt as demand, routes or populations change over time.
- Learned combinatorial optimizationGNNs and RL as solution builders for NP-hard graph problems.
Application domains
HealthHematological diagnosis, proteins and mutations
Urban mobilityPublic transit in Salvador
EducationAI literacy in public schools
Language and multimodalityEvaluation of LLMs and MLLMs
Frontier science does not have to choose between rigor and relevance.
We are a research group registered in CNPq’s Directory of Research Groups and led by CNPq Research Productivity Fellows. We publish in international AI conferences and journals, release data and code whenever possible, and train researchers at the undergraduate, master’s, PhD and postdoctoral levels.
Open science
Datasets, simulation environments and models developed at LabIA are publicly available for reproduction and reuse by the scientific community.
| Resource | Type | Description | Area |
|---|---|---|---|
| SUNT | Dataset | Reference spatio-temporal dataset on public transit in Salvador. | Graphs |
| BusEnv | Environment | Multi-agent reinforcement learning environment for urban bus systems. | Reinforcement |
| Multiple Myeloma Dataset | Dataset | Public dataset of labeled cells to support the diagnosis of multiple myeloma. | Health |
| Hemo-GNN | Model | Modeling protein activity and mutations with GNNs, with a focus on hemophilia. | Graphs |
| GNN-HemA | Model | Graph neural networks applied to hemophilia A data. | Graphs |
| MLLMs FV Evaluations | Evaluation | Evaluation experiments for multimodal language models. | Language |
Tutorials
Hands-on tutorials built by LabIA from our own datasets and research, presented at international conferences. All materials (notebooks, code and slides) are open on GitHub.
Urban mobility · Graphs
Presented · IEEE WCCI/IJCNN 2026
From Stops to Graphs
Hands-on spatio-temporal learning for public transportation
How to model a large-scale public transit system with graph neural networks using SUNT, a spatio-temporal dataset covering one year of bus operations in Salvador. From raw stop and trip data to graph construction, GNN architectures and evaluation.
- Geospatial and temporal exploration of SUNT
- Temporal models: ARIMA, RNNs, Transformers and foundation models
- GNNs (GCN, GraphSAGE, GAT) for node- and edge-level forecasting and classification
- Open challenges in graph-based mobility modeling
Instructors: Ricardo Rios, Tatiane Nogueira, Felipe Fernandes and Marcos Ferreira
Health · Computer vision
Presented · ChileUpcoming · SIBGRAPI 2026
AI for Hematological Image Analysis
Building, training and augmenting plasma cell detectors with generative AI
A hands-on AI tutorial on supporting the diagnosis of multiple myeloma from bone-marrow slide images, built on two public plasma-cell datasets created by LabIA.
- Exploring the datasets and annotations
- Fine-tuning YOLO detectors for plasma cells
- From detector to diagnosis: plasma-cell percentage per patient
- Data augmentation with generative AI (diffusion models)
Instructors: Jorge L. S. Batista Filho, Tatiane Nogueira Rios and Ricardo A. Rios
Bring a tutorial to your institution
Research groups, graduate programs, events and companies can request one of our tutorials, in person or online. Tell us a little about your group and we will get back to you to agree on format and dates.
When you submit, your email app opens with the request already filled in, addressed to ricardoar@ufba.br.
Partners
LabIA is a laboratory of the Institute of Computing at the Federal University of Bahia.
Leadership
LabIA is led by three faculty members of UFBA’s Institute of Computing. The principal investigators, Tatiane and Ricardo, are CNPq Research Productivity Fellows, with training and collaborations in Brazil, Canada, Spain and China.
Tatiane Nogueira Rios
Principal investigator · Associate Professor · CNPq Research Productivity Fellow
Intelligent decision systems under uncertainty, fuzzy logic, clustering and text mining. PhD from ICMC-USP, with research stays at McGill University and the University of Granada.
Ricardo Araújo Rios
Principal investigator · Professor · CNPq Research Productivity Fellow
Continual learning, time series, dynamical systems and chaos theory. PhD from ICMC-USP, with research stays at Université de Montréal and Hong Kong PolyU.
Islame Felipe da C. Fernandes
Coordinator · Professor
Single- and multi-objective combinatorial optimization, evolutionary computation, hybrid metaheuristics and graph theory. PhD from UFRN.
Beyond its leadership, the lab brings together PhD, master’s and undergraduate research students as well as international collaborators. Meet all our researchers.
Researchers
PhD, master’s and undergraduate research students supervised by Tatiane N. Rios, Ricardo A. Rios and Islame Felipe C. Fernandes, along with international collaborators.
- 1
- PhD student
- 13
- master’s students
- 9
- undergraduate researchers
- 2
- international collaborators
Master’s
Caroline de Souza dos Santos
Guilherme Castro Hora Fontes
Helder José Bastos Ramos
Izak Alves Gama
Jorge Santos Nery Filho
Jurgen Fink Júnior
Juvenal Bruno Andrade da Silva
Marcos Adriano Pereira dos Santos
Maíra Matos Araujo
Rafael da Costa Fonsêca
Tércio Sales Cruz
Wesley da Silva e Silva
Yves Luan do Rosário Moura
Undergraduate research
International collaborators
Click a name to open the Lattes CV or academic profile.
Research with us
We welcome students and researchers interested in AI with scientific rigor and real-world impact. Graduate admission is through PGCOMP’s calls; for undergraduate research and postdoctoral positions, contact us directly.
- Undergraduate researchUndergraduates who want to start doing AI research.
- Master’s and PhDThrough UFBA’s Graduate Program in Computer Science.
- PostdocResearchers with projects aligned with our research areas.
- PartnershipsCompanies, government and institutions interested in AI R&D.