Boulder, Colorado, United States
I'm a Research Scientist Manager at Meta where I manage a team of 6 research scientists. I operate in a TLM role where I split my time between managing and IC work such as coding and research. My main area of expertise is in sample-efficient black-box optimization and developing new methods that can be used to solve business critical optimization problems. I was previously a Sr. Research Scientist at Uber AI Labs where I was involved in building up and democratizing the use of Bayesian optimization across the company. I have a PhD in Applied Mathematics from Cornell University.
- I'm a Research Scientist Manager at Meta where I manage a team of 6 Research Scientists. My main area of focus is sample-efficient black-box optimization applied to AutoML (hyperparameter tuning, feature selection, etc.). - Our team is part of the Central Applied Science org. We develop high-impact scientific AI/ML methods that we publish in top venues such as NeurIPS/ICML. Our methods are broadly used (>1000 internal users) to drive topline impact and strategic value for Meta. - I operate in a TLM role where I split my time between managing and doing IC work. - I work closely with and maintain deep connections with tech leads in all major orgs, e.g., IG, FB, Ads, RL, and GenAI. This allows me to identify impactful areas for research.
- AutoML tech lead in the Central Applied Science (CAS) org. My main focus is identifying new research problems across the company, developing and publishing new research methods (e.g., SAASBO) that can solve these problems, and making these methods available both internally and in our open-source packages BoTorch/Ax.
- Research Scientist in Zoubin Ghahramani's lab. My research focused on high-dimensional Bayesian optimization and Gaussian processes. - I built and designed Uber's service for Bayesian optimization which had hundreds of users across the company. I proposed and led our NeurIPS 2019 spotlight paper introducing TuRBO, a scalable high-dimensional Bayesian optimization method. - Uber AI Quarterly Recognition Award in Q4 2019 (top 1%). - Uber Engineering Quarterly Recognition Award Q4 2019 (top 1%).
Project with Google Brain: - Proposed and implemented a new black-box optimization algorithm for high-dimensional problems. - Demonstrated that this algorithm outperformed Google Vizier's default. Project with Ads Search Click Quality: - Designed an ML model for ad selection. - Used this model in a live experiment and analyzed the results.
Designed and implemented surrogateopt, an asynchronous surrogate optimization framework in MATLAB.
- Developed algorithms for visualizing point clouds with billions of points. - Constructed out-of-core algorithms for shortest distance computations between a point cloud with billions of points and a geometric object.