Data Scientist · Ph.D. in Physics
I'm a data scientist specializing in A/B testing, causal inference, forecasting, and statistical modeling. Most recently, I used these skills at Meta, working on ads measurement, attribution, and incrementality. Before that, I worked at Happy Returns solving forecasting and optimization problems. I spent six years doing condensed matter research at New York University's Center for Quantum Phenomena. I also write about statistics and probability on my blog.
About me
I'm a Senior Data Scientist with a background in experimental physics, specializing in causal inference, experimentation, and statistical modeling at scale.
At Meta, I led causal measurement for third-party ads attribution products. I used propensity-score matching, synthetic controls, difference-in-differences, and randomized experiments to measure product impact and improve cross-publisher attribution. I also helped integrate view-through ads signals into third-party multi-touch attribution models, expanding measurement beyond click-based attribution.
Before Meta, I spent three years at Happy Returns (acquired by PayPal, then UPS during my tenure). I started building predictive models for shipping supply forecasting across thousands of locations and led multi-round A/B tests on customer-facing features. Over time my work shifted toward experimentation and optimization, building time-series forecasting pipelines for shipping logistics across 10,000 centers.
My PhD research at NYU's Center for Quantum Phenomena focused on current-induced spin dynamics in antiferromagnetic materials, supervised by Andy Kent. I worked with hundreds of gigabytes of electron microscopy image data and developed analytical and numerical models to extract signal from high-dimensional voltage data as a function of temperature, current, and magnetic field.
Outside of work, I write about topics I find interesting in the blog, mostly statistics, probability, and the occasional physics simulation.
Career
Technical
Writing
Notes on statistics, probability, and physics. Written for people who like thinking carefully about things.
One of the most studied problems in combinatorial optimization. The setup is simple: find the shortest route through a list of cities. Solving it exactly is provably hard. This post covers why, and walks through simulated annealing with an interactive simulation you can play with.
Two schools of thought dominate statistical inference. Using a simple coin flip example, this post walks through how each framework builds uncertainty intervals, what those intervals actually mean, and when to reach for one over the other.
What does "random" actually look like? Most people have a confident answer: scattered, spread out, no obvious patterns. But true randomness looks nothing like that. It clusters, repeats, and surprises in ways our intuitions don't expect.
When running experiments, understanding uncertainty and statistical significance matters a lot. This post explores how the bootstrapping method helps calculate confidence intervals and p-values in a concrete A/B test scenario.
Interactive
A live simulation of the 2D Ising model running in your browser: a lattice of spins that spontaneously orders below a critical temperature.
Interactive
A single magnetic moment evolving on the Bloch sphere under the Landau-Lifshitz-Gilbert-Slonczewski (LLGS) equation. Explore precession, damped relaxation, spin-transfer auto-oscillation, and field-driven switching.
Research
Physics research from my time at NYU's Center for Quantum Phenomena. Full list on Google Scholar.