Research Area
We study recommendation, ranking, and personalization methods that support relevance across Snap experiences.
Topics
Autoregressive and diffusion-based models for item retrieval, ranking, and generation.
Discrete item representations from foundation models that enable scalable generative RecSys at Snapchat.
User behavior sequences, session-based recommendation, and temporal preference modeling.
Large-scale candidate retrieval, learning-to-rank, and multi-objective optimization.
Representation learning that generalizes across domains, tasks, and data-sparse users.
LLM-based recommenders with collaborative memory and reasoning capabilities.
Explore another area