Incoming Copernicus Master in Digital Earth (GeoDSc track) | GeoAI & Advanced InSAR Engineer
I build production-grade, open-source Python frameworks at the intersection of spatial deep learning (GATs, PINNs) and physical Earth Observation invariants.
-
ps-gnn — DOI: 10.5281/zenodo.22706887 Re-engineering Persistent Scatterer (PS) target selection by reframing point cloud identification as spatial node classification via an 8-head Graph Attention Network (GAT). Status: Staged Release — Closed Beta
-
pyunwrap-insar — DOI: 10.5281/zenodo.22209219 A physics-informed neural network engine (AmbiguityNet) that resolves complex radar phase-unwrapping boundaries by enforcing discrete integer ambiguity matrices
$K$ through straight-through estimator rounding layers. Status: Staged Release — Closed Beta
-
pygeofetch — DOI: 10.5281/zenodo.22087230 A programmatic federated data-access layer spanning 24 providers with built-in pre-download burst screening (
PreflightGate) and a native, pure-Python Windows-compatible InSAR processing chain. Status: Stable — Public Production -
pygeovision — DOI: 10.5281/zenodo.22258220 An enterprise-ready GeoAI engine supporting 49 automated task pipelines, chainable preprocessing analytics, and an audited model registry (Prithvi-EO-2.0, DINOv3) driven by an LLM-backed natural language query interface (GeoAgent). Status: Staged Release — Closed Beta
- ORCID iD: 0009-0002-1376-108X
- EOCoreINT GitHub: github.com/EOCoreINT
- EOCoreINT Website: eocoreint.com
- Mexico City Subsidence Project End-to-end programmatic SBAS time-series analysis over a 502.5 GB / 67-scene Sentinel-1 SLC stack on a single machine. Mapped extreme urban subsidence vertical velocities up to −35.5 cm/year while preventing 360 GB of redundant downloads via intelligent orbital track pre-filtering.