I’m a physicist by training who leads data science teams in technology companies. My doctoral research at the University of Washington spanned several domains, from time-resolved electrostatic force microscopy and photonics to DNA pulling experiments. This diverse background taught me to apply analytical methods across different physical systems.

My early career was focused on machine learning problems. I worked on contextual bandit algorithms for fault recovery in compute systems and built distributed predictive modeling services. My perspective shifted when I joined Uber and was introduced to marketplace dynamics.

That experience sparked a fascination with the intersection of ML and marketplaces that has shaped my work since.

At Convoy, I led teams working on freight marketplace optimization. We tackled interesting problems in pricing, ranking, risk modeling, and mechanism design. I especially enjoyed building systems that matched freight shipments with carriers more efficiently and developing new bidding technologies.

Currently, I lead data science teams focused on Ads Quality, Formats, and Programmatic efforts at Pinterest. I manage over 30 scientists spanning multiple disciplines. My work centers on causal inference, experimental design, and evaluating machine learning systems in marketplace environments. I’m particularly interested in measuring how interventions propagate through complex networks over time.

Working at marketplace companies has taught me a lot about how markets actually function. It’s been interesting to see how my physics background helps when thinking about these complex systems. There’s a surprising amount of overlap in how you approach both types of problems.

I enjoy figuring out what really matters in complex systems with lots of data. Beyond just running experiments, I’m interested in understanding if changes actually make a meaningful difference in the real world. It’s easy to find statistically significant effects, but much harder to find ones that truly matter.

I started this blog to share my thoughts on science and coding problems. It’s mainly a space for ideas that bridge different fields and perspectives. I find that the most interesting solutions often come from connecting concepts across domains like physics, economics, and machine learning.

When not working with data, you’ll find me tending to a smoker. The process of making a proper brisket requires patience, attention to subtle changes, and respect for fundamental principles. Not unlike good science.