I am interested in accurate, calibrated Earth monitoring at global scale. Unlabeled Earth observation is basically unlimited, but costly field measurements are scarce and spatially clustered. Calibrated inference at high resolution over the whole Earth is intractable, and approximations that do scale well degrade uncertainty in ways that are often not well understood. My work develops amortized machine learning inference methods, analyzes where they fail, and applies them to carbon and biodiversity monitoring where downstream decisions often rely on maps with underquantified error.
I'm currently a PhD candidate at Cambridge's Energy and Environment Group advised by Srinivasan Keshav, where I also work on the Tessera foundation model. Aside from Cambridge, I also work with the Society for the Protection of Underground Networks on understanding fungi better so that we can protect them. I previously did my masters at UIUC and undergrad at Vanderbilt University.