Experience

  1. Applied Scientist, Prime Machine Learning Team

    Amazon
    • Produced a series of reinforcement learning algorithms in Python to efficiently identify the optimal characteristics of prime membership advertisements through strategic sequential testing
    • Adapted these models to complex data settings, including sparse signals and heteroskedastic noise
    • Created realistic simulation structures to test best-advertisement identification algorithms (“best-arm identification” in the multi-armed bandit literature) against A/B testing and other competitors in high-dimensional data contexts
  2. Applied Scientist, Sponsored Products

    Amazon
    • Designed a reinforcement learning algorithm to better select advertisements sourced by a variety of machine learning algorithms using query features and partially observed customer behavior signals
    • Implemented the infrastructure for evaluating the advantages of modifications to advertisement sourcing algorithms on downstream outcomes in Spark
    • Ran off-line experiments testing the utility of my algorithm over a week of Amazon search query traffic and found that it was consistently capturing more than 3% of the high-quality ads missed by the current method

Education

  1. PhD in Statistical Science

    Duke University
  2. BA in Mathematics

    Pomona College