• Own the Machine Learning-driven ads bidding end-to-end

• Develop new features and improve tree and DNN models to improve click-through rate and conversion rate

• Built new bidding products (target ROAS/Spend/Acquisition etc) on search ads, video ads, and offsite ads.

• Drive bidding optimization improvements through Feedback/ROAS Model/Selective Serving etc

• Increased daily traffic and revenue by 21% and 8%, respectively, by implementing recommendation engine to individualize user home feed, item notification, weekly digest, and targeted campaigns.

• Reduced processing time of machine learning-based fraud detection system from 35 mins to 15 secs to discover scam and content abuse by training decision tree and logistic regression models.

• Processed 5 million daily requests with less than 200 ms average latency from marketing campaigns and in-app advertising by delivering user profile, more than 200 user segments, and 3500 popular tags.

• Acquired 1.3 million new users by performing A/B testing, friend invitations, and gamified referral bonuses.

• Improved 7% of weekly orders by applying collaborative filtering and Apriori algorithms to recommend products on homepage, cart and checkout pages.

• Increased 13.7% of daily page views by utilizing click-through rate prediction to Women’s Clothing, Shoes and Bags, Wedding and Events, and Consumer Electronics categories.

• Compressed ETL and report generation time from 90 mins to 5 mins by building a distributed streaming data pipeline on AWS to replace local on-premise Hadoop clusters.

• Developed map-reduce jobs to process terabytes of data on Hadoop clusters for financial customers.

• Migrated jobs from mainframes onto Hadoop clusters to save customers 30% of annual cost.