Research Associate · Multimodal Vision Research Lab · WashU
computer vision, geospatial AI, and retrieval at planetary scale
I build models and systems that reason about the Earth: location encoders that learn where to spend their capacity, and retrieval methods that find where a photo was taken when the gallery is the whole planet. I work in Nathan Jacobs’ lab at Washington University in St. Louis, and I’m Chief Scientist at 1088 Advisors, a small higher-ed strategy consultancy.
Before this: three years as a systems engineer at an automotive radar startup, a stint in healthcare data engineering, and several years in neuroscience labs.
A location encoder built from learnable spherical Voronoi partitions. Instead of a fixed basis that treats open ocean and a growing city the same, the cells themselves migrate to where the data needs resolution.
Finding where a photo was taken by retrieval usually means comparing it against every tile of a planet-scale gallery. This work organizes the gallery hierarchically and learns which regions are worth refining, matching exhaustive search at a fraction of the encoding cost. Name and details withheld while the paper is under review.
A link-budget and scan-configuration tool for digital-code-modulation radar. Predicts SNR across range, Doppler and angle for any antenna and scan preset, with a Dash front end that engineers used without touching code.