I study how inequality accumulates across education, careers and health — at the scale of whole populations.
Data Science Fellow and New Map of Life Fellow, Stanford University
I work with Mitchell Stevens and David Rehkopf at Stanford. I received my PhD in computer science at CU Boulder in 2025, advised by Aaron Clauset and Dan Larremore and supported by an NSF Graduate Research Fellowship, alongside a Master's in education policy. I have been a Data Science Fellow at the U.S. Census Bureau and a graduate research assistant at the National Center for Women & Information Technology. Before the PhD I was a research engineer at IBM Research, and studied computer science at Indiana University, where I built a system to detect dyslexia in children's handwriting. Full history on the CV.
Life-course research has studied turning points — events that durably redirect a trajectory — since at least Elder's work in the 1980s, and has developed a range of ways to identify them. These projects look for turning points at scale in longitudinal records: which diagnoses redirect a patient's disease trajectory in primary care, and which events in early adulthood shift a life trajectory, for whom.
A summer in the Census Bureau's Enhancing Health Data group, developing a community-level measure of the risk that access to healthcare is declining.
Research on gender in higher education has tracked representation and attrition since the 1970s and documented the role of departmental climate. These papers add large-scale evidence from the employment records of nearly every professor at US universities, paired with survey and interview data: women leave faculty positions at higher rates than men at every career stage, tenure included, and faculty who leave more often cite their workplace than anything pulling them elsewhere.
Analog, non-volatile memory can perform the matrix multiplications in a neural network directly in hardware, at a cost in precision. This work measured and narrowed that cost for transformer-based models on phase-change memory, largely through noise-aware training.
Dyslexia affects roughly one in five people in the US, and benefits from early detection, but screening tests are applied inconsistently and classroom teachers are not generally trained to administer them. These two workshop papers contributed handwriting datasets from children with and without dyslexia, a computer-vision classifier intended to run alongside existing screening, and a follow-up examining whether such a system should be used at all.