About

Steven Scott specializes in Bayesian statistics, forecasting, multi-armed bandit experiments, latent variable modeling, hierarchal models, finite mixture models, hidden Markov models, and model selection. He is interested in data augmentation methods and Markov chain Monte Carlo. He has applied these methods to problems in educational testing, network security, causal inference, biometrics, web browsing, e-commerce, and medical applications.

Aside from being a statistician, Steven Scott is also a researcher, educator, senior analyst, and public speaker. He worked as an Assistant Professor of Statistics at University of California’s Marshall School of Business from 1998 to 2007. He also served as the Director of Statistical Analysis at Capital One Financial Corporation from 2007 to 2008.  He worked with Google from 2008 to 2018 as a Senior Economic Analyst and Director of Statistics Research.  

While working for Google, he made significant contributions in the Thompson Sampling algorithm for multi-armed bandits that gained substantial traction both inside and outside Google. Steven Scott also wrote the code for the mathematical back end of the system that served multi-armed bandit experiments for third-party website optimization.