Research
My research spans medical imaging and natural language processing, with a focus on deep learning, computer vision, and efficient transformer-based models. The following are some highlights that describe my lines of research.
Medical Imaging Processing
- Principal researcher at MICLab (Medical Imaging Computing Laboratory)
- Analytical AI for medical imaging: automated segmentation and classification in clinical images (CT, MRI, and diffusion)
Examples
- MEDPSeg: hierarchical multitask segmentation of pulmonary structures and lesions in chest CT
- e2dhipseg: extended 2D consensus hippocampus segmentation in brain MRI
Healthcare Informatics
- Vision-language models combining clinical text and medical images for triage, classification, and decision support
- Efficient and privacy-preserving deployment: on-premises inference, quantization, and local integration with hospital PACS
- Agentic smart PACS implementations embedding local models and tool-augmented agents into radiology workflows and multimodal imaging archives
Examples
- PrioScan: offline deployment of VLMs for automated chest X-ray prioritization at HCPA
Natural Language Processing in Portuguese
- Natural language processing in Portuguese, including pre-training and evaluation of transformer models
- Transformer architectures and efficient small language models
- Plain language and applications aligned with Brazil's National Plain Language Policy (opportunities)
Examples
- PTT5: T5 model pre-trained on Brazilian Portuguese
Open science
- Open-source software, pre-trained models, and reproducible pipelines released with publications
- Public benchmarks and shared datasets to support comparison and reuse by the community
- Tools designed for practical use beyond the paper—demos, packages, and ready-to-run implementations