Geospatial ML researcher. I build graph neural networks for Earth observation: terrain correction from sparse spaceborne LiDAR, physics-informed RF propagation over heterogeneous terrain, and mobile-device geolocation. M.Sc. student in Defense Engineering at IME (Brazil).
Every repository here is a reproducibility package for a manuscript: code, stamped results and the audit scripts that regenerate every reported number.
- Published: Seelig & Salles, ARPIA: a multi-scale heterogeneous GNN for mobile-device geolocation, Ad Hoc Networks 189, 2026.
- Tools: Python, PyTorch, PyTorch Geometric, GDAL/rasterio, QGIS.
- Contact: s33l1g@gmail.com · LinkedIn