About Me
I am a fifth-year Ph.D. candidate in the [Department of Statistics](https://stat.uw.edu/) at the University of Washington, advised by Prof. [Alex Luedtke](https://www.alexluedtke.com/). My research lies at the intersection of causal inference, optimal transport, nonparametric statistics, and machine learning. At UW, I am fortunate to work with Profs. [Zaid Harchaoui](https://sites.google.com/uw.edu/zaid-harchaoui/main) and [Soumik Pal](https://sites.math.washington.edu//~soumik/). I also collaborate with the [Abrahms Lab](https://www.abrahmslab.com/) and the [eScience Institute](https://escience.washington.edu/) on developing AI tools for ecology.
I completed my undergraduate degree in Mathematics at the Indian Institute of Technology Kanpur, with a minor in English literature. At IIT Kanpur, I worked with Dootika Vats on output analysis and convergence diagnostics for Markov chain Monte Carlo and importance sampling.
Contact me at medhaaga [at] uw [dot] edu.
News
May 2026
Looking forward to attending the Optimal Transport + Optimization Workshop in Les Diablerets, Switzerland.
Winter 2026
Looking forward to being a Visiting Scholar at Harvard Medical School.
July 2025
Our paper Leveraging machine learning and accelerometry to classify animal behaviours with uncertainty was published in Methods in Ecology and Evolution.
June 2025
Looking forward to attending the workshop on Wasserstein Gradient Flows in Math and Machine Learning at the Banff International Research Station.
Feb 2025
Gave a talk at the IFML Mathematics of Deep Learning Workshop in Austin, TX. [Slides]
Feb 2025
Gave a talk at the UW Data Science Seminar, titled Revealing the Hidden Lives of Cryptic Carnivores with Machine Learning and AI. Check out the YouTube video! [Slides]
Selected Publications
Accepted at International Conference on Machine Learning (2026)
A distributional treatment effect based on Sinkhorn divergence, with efficient first-order estimation and second-order null inference for testing equality of counterfactual outcome laws.
Published in Methods in Ecology and Evolution (2025)
An open-source pipeline combining accelerometry, 1D convolutional neural networks, class rebalancing, temporal smoothing, and conformal prediction for uncertainty-aware behavior classification.