Benu Changmai

Meta GenAI Discovery Team (current)

• Developing GenAI-powered entry points, enabling user interaction with MetaAI and its capabilities
• Helped expand MetaAI to support 7 international languages on MetaAI entry points
• Currently leading efforts to build new sources for relevant prompt suggestions and improve personalization by developing the ranking module for MetaAI prompts

Facebook Video Ranking Team

• Built the ML foundations for Video Ranking and Recommendations for growing Video (reels) creators and their content on FB
• Improved creator metrics such as Producer DAU (+7.3%), manifold engagement quality, and cold content breakout (+20%)
• Built critical infrastructure for content retrieval, focusing on efficient indexing, query processing, and real-time video retrieval
• Enhanced personalization through improved early-stage ranking by user interest profiling and generating rich content representations

Alexa AI Dynamic Routing

• Developed infrastructure and software solutions for ranking Alexa skills which serve voice requests of Alexa voice assistant customers
• Designed and built platform for model learning of rule- based ranking policies resulting in enhanced model performance and reduced reliance on static rules (Alexa AI)

On-device AI team

• Worked as a research engineer dedicated to bringing artificial intelligence to software solutions in Android Platform Division
• Successfully commercialized smart feature called AliveSMS for India-based Samsung smartphones which intelligently extracts important user-actionable information from SMS
• Built on-device light-weight neural models (CNN, RNN, LSTM, ResNet) and NLP solutions (NER, LDA, classification, natural language parsers) as API services for the Bixby personal assistant in Samsung mobile devices
• Built on-device Knowledge reasoning inference engines for use in personal recommendation systems (part of Bixby assistant)