As of March 2017, I work with the Streaming Science and Algorithms group at Netflix, where we use a combination of statistical modeling and experimentation to optimize the streaming user experience for our customers around the globe.

Prior to joining Netflix, I spent two years as the Principal Statistician for the Natural Perils Research Team at Insurance Australia Group in beautiful Sydney, Australia. At IAG, I led natural perils risk modeling projects with insurance and reinsurance applications, including the development of claims-based pricing models. I also worked extensively on the on-going reinsurance settlement of losses arising from the 2010-2011 Christchurch Earthquake sequence, “the most complicated insurance settlement program experienced anywhere.”

Prior to joining IAG, I was an Assistant Professor in the departments of Statistics and Meteorology at Penn State University, where I conducted fundamental research in statistical climatology. Much of my academic research involved inferring geophysical processes, such as surface temperatures, from numerous sources of data that each have different uncertainties and different relationships with the underlying process. The observations, as well as the geophysical processes, typically display spatial and temporal dependencies, which motivates the hierarchical statistical approach that is a common thread to my research.  Such models allow for the construction of scientifically-informed, space-time relationships for the geophysical process under analysis, and for the dependence of each type of observation on that process. Bayesian inference then allows for a complete treatment of uncertainty, and posterior samples of the geophysical process can be used to answer a wide array of scientific questions. These same driving concepts now underpin my work at IAG.

My graduate education was completed in the Department of Earth and Planetary Sciences at Harvard University, where I worked under the peerless Peter Huybers. My dissertation involved the development of a simple hierarchical model to assimilate various surface temperature proxies (ice cores, tree ring widths and densities, etc.), along with the modern instrumental record, to reconstruct the spatial pattern of past climate. This work was novel in the paleoclimate literature for specifying a parametric space-time covariance model for the surface temperature process, and then specifying both instrumental and proxy observations as functions of this latent process. I then spent time as a postdoctoral fellow at SAMSI, where I participated in the Program on Space-time Analysis for Environmental Mapping, Epidemiology and Climate Change and helped lead the Paleoclimate Working Group, and with the IMAGe group at NCAR.

You can find an updated publication list at Google Scholar.