Research Scientist,Meta Recommendation System Core modeling

About the job

The Meta Recommendation Core Modeling team is a central group within Meta focused on building and advancing the core machine learning models that power recommendations across Meta’s products. Our team (MRS Representation Learning)’s mission is to invest in Cross-Meta (ads+organic) joint modeling technologies and unified tech stack to build omni user representation and foundation models to improve Meta’s recommendation systems and maximize business and user value, which is a critical stepping stone in the Meta Recommendation System vision. The team is building advanced recommendation ML models through redesigning model architectures and developing data learning technology at a state-of-the-art level to support a massive user base with accurate predictions and high queries-per-second (QPS) performance.

Responsibilities

Conduct original research in a recommendation/ranking system Design and execute experiments to develop and validate novel ranking modeling techniques, training objectives, and evaluation methodologies Develop scalable approaches to language model pretraining, fine-tuning, and instruction following that advance the state of the art Analyze model behavior, failure modes, and generalization properties to generate actionable research insights Translate research findings into high-quality publications submitted to top-tier venues such as ACL, EMNLP, NeurIPS, ICML, and ICLR Collaborate with cross-functional partners including engineers, product managers, and data scientists to apply language research to Meta's AI systems and products Contribute to open-source releases and reproducible research artifacts that benefit the broader NLP research community Mentor other researchers and engineers on the team, sharing technical expertise and providing feedback on research direction and execution Identify and drive new research directions aligned with the team's long-term goals, proactively scoping projects and building stakeholder alignment Use AI-augmented workflows to expand research productivity, accelerate experimentation, and explore cross-disciplinary problem spaces