At HAUZ.ai, I’m the main author of an ArcGIS-to-Neo4j pipeline that discovers Florida GIS portals, classifies map layers, and loads them into a knowledge graph. I designed its graph model and reduced Neo4j timeouts from 36 to 8 in a local benchmark.
I also developed a geospatial ingestion system for Clairvant.io, including a Python/GDAL map worker that loaded 3,193 Puebla layers from public INEGI data with no sync failures.
For HAUZ.ai, I built a browser-based 3D house viewer and a custom NVIDIA Omniverse Kit app that streams an OpenUSD digital twin over WebRTC. I also prototyped GraphRAG chat and a local-LLM agent for the viewer.
Earlier, I developed software and interactive experiences at The Line Production Lab, TROOPSMX, Ideography, and as a freelance developer. My projects include MICI, a geospatial AI platform for Mexico, and Cademy VR, a WebXR forklift trainer.
