Sarah Shi is a petrologist and data scientist working at the intersection of geochemistry and machine learning. She holds a B.A. in Earth and Environmental Science from Columbia University, where she also worked as a Data Science Fellow in the Geoinformatics Research Group at the Lamont-Doherty Earth Observatory. She completed an MPhil in Earth Science at the University of Cambridge as a Euretta J. Kellett Fellow, and is currently a Ph.D. student in Earth and Planetary Science at the University of California, Berkeley. Her research centers on extracting quantitative information from mineral compositions and on making those extractions transparent, reproducible, and uncertainty-aware. She develops open-source tools to address this directly: mineralML performs probabilistic mineral classification from oxide chemistry using neural networks, returning calibrated confidence scores so that uncertainty is explicit rather than hidden, and PyIRoGlass applies Bayesian inference to volatile quantification in melts. Sarah is interested in how statistical and computational methods can make geological interpretation more open, reproducible, and honest about what we do not know — values directly relevant to responsible resource characterization.
Type and press enter to search
