This blog post discusses the development of a destination recommendation model designed to help users explore travel options on the Airbnb platform. The authors address various challenges in modeling user intent based on their actions and preferences, utilizing a framework inspired by language modeling. Key features include balancing the behavior of active and dormant users, integrating geolocation information, and applying multi-task learning. The model has applications in enhancing user engagement through autosuggestions and abandoned search notifications, aiming to drive more bookings.