During my seven years at Yelp, I worked almost exclusively on high-value ads systems and ETLs. I initially contributed to the cost-per-click ad bidding system, helping to build features that increased budget fulfillment while lowering advertiser costs. In later years, I switched to supporting generation of mission critical data sets used to drive revenue decisions, exploratory analysis, and downstream ML applications. This involved rewriting large PostgreSQL queries to run as much as ten times faster and rebuilding legacy Map Reduce jobs to run on Apache Spark.

Implemented a machine learning classifier for predicting the sentiment polarity of Twitter messages; designed the classification feature set, implementing and adapting strategies from related research projects; conducted experiments using example data sets from the SemEval-2014 workshop