• Technical lead for the ground truth labels used for training and evaluating the age and location machine learning models on Facebook and Instagram. Set the direction and goals on ground truth strategy for cross-functional teams of data scientists and software engineers.• Developed and implemented methodologies to reduce the cost of label collection, improve the accuracy and representativity of labels, and estimate the performance of ML models on users in countries without available labels.• Co-developed the Ground Truth Maturity Framework to assist teams across the company withassessing and improving the data used in their machine learning models.

• Built GBDT classifiers to predict users' education information by leveraging their social graphs.

• Implemented a more efficient estimator for certain metrics which resulted in reduced false negative rates in AB tests• Trained a text classifier for predicting whether a question is clickbait• Improved personalization of topic recommender for new users by incorporating external data sources

• Presented papers on the response of foreign exchange markets to macroeconomic news announcements and the potential for different arbitrage opportunities• Analyzed the time-dependent effect of signed volume on currency exchange rates with a large dataset from EBS• Examined the potential for triangle arbitrage in the FX market from 2001-2013 using C and R

• Used Excel and Python to create a data set from the results of an field experiment on telemarketers• Conducted an econometric analysis on the effects of various payment incentives on telemarketer behavior and success rates using Stata and the generated data set• Assisted in conducting field experiments on ultimatum games and risk aversion• Reviewed and critiqued experimental economics papers• Collaborated with other RAs to discuss methods and increase efficiency

• Explored measure-theoretic probability with the guidance of a graduate student mentor• Wrote an expository paper on the convergence of random series and martingales using LaTeX