Recommendation systems largely rely on engagement (clicks, likes, comments) for ranking but that doesn't tell the full story of what users value. Currently leading an effort at Meta to leverage what user's say what they want to see more or less of in recommendations across Facebook products.

Led the work on aligning Facebook's recommendation and ranking systems with core human values like connection, empathy, care, inspired and being informed. The work involved research in deep content understanding across text, image, audio, ocr to predict how well the content aligns with what a user values.

* Led an effort to improve the speech recognition stack in Facebook Assistant and bring it up to state of the art. Built personalized name recognition that could learn and adapt to how an individual pronounces names of their friends.* Built a chat-bot based on a large language model trained on transformer architecture

Technical lead for Messenger product infrastructure and led the design and development of:* Facebook Realtime Infrastructure: A pub/sub system for routing realtime messages* Facebook Push Notification System: A push system for delivering notifications for Facebook family of apps on Android* Omnistore: a data store for mobile devices that is sync'd in real time with cloud and other devices

Managed the Core Systems group and has led the design and development several systems at Facebook, including:* A fault tolerant distributed key/value store* Facebook wide distributed coordination and locking service* Facebook’s service discovery and shard management systems, to bring automatically failover and load balancing to services at Facebook

Led the database management team in SQL Server. SQL Azure is SQL Server’s cloud database platform that supports thousands of databases across hundreds of machines in a cluster. The database management in SQL Azure is responsible for keeping an inventory of all the databases in the system, providing metadata for name resolution and connection establishment.

Evolving SQL Server into highly available, scalable distributed database system. Responsible for technically leading the design and implementation of the data layout management and data high availability components in this system.