Welcome!
I'm a fourth-year PhD student at MIT in the Laboratory for Information and Decision Systems (LIDS) and the Institute for Data, Systems, and Society (IDSS), advised by Professor Tamara Broderick.
My interests lie broadly in statistics and machine learning, particularly causal inference and interpretability. More recently, I've become especially interested in developing machine learning methods and using AI tools to improve healthcare, with applications to disease understanding, diagnosis, treatment, and clinical decision-making.
Before MIT, I did a master's in mathematical statistics at Cambridge (Part III), and before that I studied applied math and economics at ITAM in Mexico City, where I spent a few years running the analysis for field experiments on gender-based violence and misinformation.
I enjoy spending time outdoors: cycling, swimming, being in the mountains, and lately, rowing.
The best way to reach me is by email. You can find my address on my CV.
Research
My PhD work is in statistics and machine learning: how to explain differences between populations, causal inference, and interpretability, with a growing focus on healthcare. Before that, I analyzed randomized field experiments on media, misinformation and gender-based violence.
The Oaxaca–Blinder decomposition splits a gap between two groups into differences in covariates and differences in outcomes given covariates. It requires picking one group as the reference, and we show that choice can flip the substantive conclusion, even with flexible models like a pretrained transformer. Report both directions.
Large construction projects link many contractors, so one failure can spread. We model the contractor network as a directed graph, extend the Friedkin–Johnsen model to simulate how failures propagate, and find about 2% higher exposure for an insurer once network effects are included.
Extending the Kitagawa–Oaxaca–Blinder decomposition with nonlinear tools like functional ANOVA or accumulated local effects seems natural. We show these can blame outcome differences on factors that are identical in both populations, and characterize when that happens.
A randomized evaluation in Egypt of media campaigns aimed at women facing gender-based violence during the pandemic. I led the statistical analysis.
An experiment in Bolivia on whether digital literacy training and fact-checks help people tell true news from false.
The proposal behind Mexico's business cycle dating committee, which INEGI formally created in 2021.
α–β authors listed alphabetically.
Experience
Teaching