Beyond Precision: Building Trust and Safety in AI-Powered Content Recommendation
About This Session
The future of AI-powered content recommendation demands more than just technical precision—it requires building systems that users can trust. I will share frameworks and methodologies for evaluating recommendation quality beyond traditional metrics, incorporating dimensions of transparency, accountability, and user agency. Through case studies from the industry, attendees will gain insights into effectively navigating trade-offs between innovation and responsible deployment, establishing appropriate human oversight mechanisms, and designing recommendation systems that not only understand content but respect the complex human values and preferences they serve. This session offers practical guidance for organizations seeking to harness the power of multimodal LLMs while maintaining robust ethical standards in their recommendation practices.
Speaker
Aashu Singh
Senior Staff Software Engineer - Meta Platforms Inc
Aashu Singh is a Senior Staff Software Engineer at Meta with over 9 years of experience, specializing in Multimodal Large Language Models for content understanding across Facebook and Instagram recommendation systems.
At Meta AI, he leads initiatives that use multimodal LLMs to enhance relevance in recommendation systems, improving how machine learning interprets content across modalities for personalized experiences. He's co-authored key publications including "Transfer between Modalities with MetaQueries" (2025) and "CompCap: Improving Multimodal Large Language Models with Composite Captions" (2024).
Previously, he contributed to Meta's advertising ecosystem, developing models for Ads Retrieval, Ranking, and Dynamic Ads for hyper-personalized user experiences
At Meta AI, he leads initiatives that use multimodal LLMs to enhance relevance in recommendation systems, improving how machine learning interprets content across modalities for personalized experiences. He's co-authored key publications including "Transfer between Modalities with MetaQueries" (2025) and "CompCap: Improving Multimodal Large Language Models with Composite Captions" (2024).
Previously, he contributed to Meta's advertising ecosystem, developing models for Ads Retrieval, Ranking, and Dynamic Ads for hyper-personalized user experiences