Accomplishments
Other notable achievements include:
Over the course of several years, I was quickly promoted to a Senior Scientist position. Recognized for my successful efforts to grow YouTube Shorts, I played a pivotal role in the growth and development of the product.
I’m passionate about statistics, and am always eager to help peers and colleagues with their work. Recently, I’ve been authoring the core statistical analyses for Pando Pooling, a startup aimed at pooling professional athletes together (with similar projected earnings) such that they may lower their individual career risk. We collected data on over fifty thousand professional baseball players, alongside their career earnings and player-statistics. These data are then used to construct models which estimate a player’s (remaining) lifetime earnings, as well as the likelihood of making the Majors or reaching arbitration.
As first “data scientist” at Cardinal, I have built out the core product(s), authoring the feature engineering and modeling packages. The main implications are improvements above and beyond the prototypical publication that defines "Cost Blooms" and how to mitigate them ( Predicting patient 'cost blooms'). Specifically, I expanded our feature set for each patient from 800 variables to over 80k, which improved model performance by 50% when used in conjunction with a more generalizable ML architecture. I also dropped the memory footprint and compute time by orders of magnitude through use of sparse matrices and a mind toward performant algorithms. I’ve extended their models to work for arbitrary labeling functions, e.g. predicting a patient’s risk profile or orthopedic joint surgeries; this helps us realize more value with new clients.
I also worked at Lawrence Livermore National Laboratory for a summer, implementing distributed MCMC in Scala and Spark according to several different methods recently published, e.g. Neiswanger et. al..
I worked for several years as a Research Analyst with The Brattle Group, where I became passionate about computing in R. I worked closely with Armando Levy to estimate the damages from lost human use of select natural resources used for recreation, for the purpose of supporting the expert testimony of Dan Mcfadden. Along the way, I authored a sequence of over a half-dozen R presentations, informing RA’s of more efficient ways to tackle their daily workflow; the content led to presentations such as a how to on regular expressions.