San Francisco, California, United States
I'm a machine learning research engineer at Netflix. Previously, I was one of the founding members of Pinterest Monetization team. And before that I was at Microsoft working on Bing Ads Core Relevance. For deep learning tutorial series visit my medium blog: https://medium.com/@ardendertat For how to implement a search engine tutorials refer to my blog: http://www.ardendertat.com/2012/01/11/implementing-search-engines/ For programming interview questions check here: http://www.ardendertat.com/2012/01/09/programming-interview-questions/
Leading several high-impact, cross-functional initiatives across product, ads and personalization: - Member Ads Targeting: developed a causal model to target Netflix members with ads. - Acquisition Ads Targeting: built an incrementality based algorithm to target non-members with ads. - Sole owner of the models that guided $100M member and $1B acquisition ad spend on Youtube/Meta/Google. - Account Sharing: launched the policy that generated multi-$B of revenue over several quarters. - Messaging: architected a deep learning model to personalize the content and timing of push/email notifications. - Evidence: designed and implemented image/video/text recommendation systems that personalize title artwork. - Total contribution on the order of a 1% improvement in streaming time for all Netflix members.
- One of the founding engineers of Pinterest Monetization team. - Tech lead of Ads Selection and Relevance areas. - Designed and implemented query expansion service for search ads. Improved revenue and clickthrough rate. - Built the first deep learning model in the ads team. Pioneered the usage of Tensorflow and Keras. - Rewrote ads indexing and selection. Computed several high-quality signals, increased precision and recall. - Built machine learning models to calculate query to ad relevance and filtered out irrelevant ads. - Added new features to the click prediction model and improved the accuracy of user to ad relevance model. - Tech lead of audience targeting feature which enabled unique targeting options (press coverage). - Several double-digit revenue, coverage, and engagement gains. Owner of multiple high-priority work streams. - Overall contribution in the order of $50M, 20% engagement gain, 60% relevance gain.
- Enriched ads with new annotations to improve click yield. Increased Bing’s revenue by 1.5%. - Improved CTR by 3% and dwell time by 2% without introducing quick-backs. - Developed tools for training data validation, feature comparison, ranker calibration, and KPI calculation. - Improved Bing’s relevance metric by more than 8% with less than 0.5% revenue loss - Worked on query specific, document oriented, and query-document cross features. - Developed machine learning models to rank candidate ads for a given query real-time in production.
- Worked jointly with Index Generation and Relevance on landing-page related features. - Added a new stream to the main production index. Saved 25% space without relevance loss.
- Developed and implemented a website to monitor the live traffic of Core Network. - Modeled and documented the data traffic flow within the 3G Network. - Installed an IP service optimization solution to the 3G Network.