Vibhor Rastogi

Engineering Manager at Facebook

Vibhor Rastogi is a seasoned professional in the field of computer science and machine learning, currently serving as an Engineering Manager at Facebook (now part of Meta).

Education

He holds a Bachelor of Technology (BTech) in Computer Science from the Indian Institute of Technology, Bombay, which he completed from 2001 to 2005.
He earned his Doctor of Philosophy (PhD) in Computer Science from the University of Washington, specializing in privacy-preserving machine learning using Differential Privacy, from 2005 to 2010.1 2

Professional Experience

Engineering Manager at Facebook (Meta): Since June 2017, Rastogi has been supporting machine learning and privacy teams for business messaging across Facebook apps, including Facebook Messenger, Instagram Direct, and WhatsApp. His work involves developing novel deep learning models, building large-scale retrieval and ranking systems, and ensuring user privacy through state-of-the-art anonymization techniques. His efforts have significantly impacted Facebook's revenue and user experience, leading to billions of dollars in incremental revenue, an 80% increase in conversions on messaging threads, and a 75% reduction in scams.1 3
Senior Staff Data Scientist at Twitter: From December 2013 to May 2017, he worked on various aspects of Twitter ads, including targeting, prediction, and auctions. He led several large initiatives that resulted in hundreds of millions of dollars in incremental revenue.1
Research Scientist at Google: From June 2012 to December 2013, Rastogi worked on using long-term signals for ads personalization at Google Research.1
Research Scientist at Yahoo: From October 2010 to June 2012, he was part of the database group at Yahoo! Research, focusing on large-scale entity matching.1

Specializations

Rastogi has a strong background in machine learning, privacy-preserving technologies, and large-scale data systems. His work often involves balancing business goals with user privacy, leveraging advanced techniques like differential privacy.1 3