About Me
I am a currently a researcher associated with the PUC-Behring AI Institute and soon will join the Departamento de Informática at PUC-Rio as a full professor.
Previously I was a researcher and manager at IBM Research Brazil, where I led multiple applied artificial intelligence research projects in areas spanning from subsurface characterization for oil&gas and mining, to earth observation, weather, and climate problems.
For a full academic and professional history, you can view my Curriculum Vitae.
Research
My main area of focus is developing and analyzing machine learning methods that use geospatial data for climate & sustainability applications. Some research topics of current interest are:
- Robustness: How can large models pre-trained on unlabeled geospatial data capture the different spectral, spatial, and temporal nuances required for various types of applications?
- Active Learning and Annotation Efficiency: How can we direct limited labeling efforts to maximize model performance?
- Interpretability and Generalization: How can geospatial priors and physical models improve transparency and robustness?
- Benchmarking and Impact: How can evaluation metrics reflect real-world impact and enable a fair comparison between models?
- Local vs. Global Models: Should we strive for a single global models or multiple local specialists, and can we automatically identify the scopes for transferability?
- Multimodality: What are the benefits of multimodal pre-training for tasks that include multiple modalities, and what are the potential ways to incorporate additional data during the fine-tuning phase?
For a full list of papers please see my Google Scholar.
Teaching
-
Artificial Intelligence for Geospatial Data (Graduate level)
Concepts, models, and applications of the field known as GeoAI — artificial intelligence for geospatial data — preparing the student for future research on specific topics within the field.
-
Introduction to Artificial Intelligence (Undergraduate level)
Foundational course in AI with a particular emphasis on intelligent agents, problem-solving, logic, and automated planning.
Contact
Email: biancaz@gmail.com