Welcome to my personal page. Currently, I help build Eppo, an experimentation platform to empower the curious and entrepreneurial as the Head of Statistics Engineering; my time is roughly equally split by doing statistics, engineering and product work.
Previously, I led the Core Representation Learning team at Stitch Fix. Our team builds multiple recommendation systems that help understand client perferences across multiple dimensions such as style and size and form the backbone of all recommendations across Stitch Fix products.
I obtained my PhD at Stanford working with Ramesh Johari. My interests are at the interface of statistics, mathematics and computer science.
Experience
-
Eppo, 2021 - Now
The statistics engineering team focuses on combining statistical rigour with an intuitive product that empowers everyone to analyze experiments with confidence. Projects include work on sequential confidence intervals, implementing CUPED, and automated power analysis.
-
Stitch Fix, 2018 - 2021
On the Algorithms team at Stitch Fix I have worked on a variety of machine learning, statistics and opimtization problems. These include
- Multiple algorithms based on latent embeddings: similar item search, diversifying recommendations, and generating outfits
- An experimentation framework to deal with spillover effects from inventory constraints
- Solving large scale optimization problems (distributed in PySpark, and with state of the art commercial solver)
- Creating a production system that serves Style Shuffle quizzes to clients in real-time
- Proposing, running, and analyzing multiple experiments
among others.
-
Stitch Fix internship, Summer 2014 and Summer 2015
Stitch Fix is reinventing the retail industry through innovative technology. During my two summers on the Algorithms team at Stitch Fix, I worked on improving the recommendation engine that is used by stylists, along with tinkering on some side projects.
-
HP Labs internship, Summer 2013
HP Labs is the research division of Hewlett-Packard. I worked under supervision of Rob Schreiber and with fellow intern Austin Benson on fault tolerance for the next generation of super computers. We demonstrated that we can make numerical methods resilient to silent errors at negligible cost by using mathematical properties of the methods. This has led to the publication of Silent error detection in numerical time-stepping systems.
Education
-
PhD Computational and Mathematical Engineering, 2018
Advised by Ramesh Johari
Stanford University
-
MASt Mathematics, 2012
University of Cambridge
-
BSc Econometrics and Operations Research, 2011
University of Groningen