Large Scale Pathways Recommender Systems (PaRS) at Verizon

Explore the development and implementation of a novel multi-task learning recommender system called Pathways Recommender System (PaRS) at Verizon in this 31-minute conference talk. Learn how PaRS leverages the Ray ecosystem to overcome challenges in digesting heterogeneous data, accomplishing various tasks, and efficiently training large deep learning models at scale. Discover how this system enhances customer experience by better understanding behavior across multiple channels, including website visits, store interactions, and customer service engagements. Gain insights into the system's ability to handle abstract forms of data, generalize across multiple recommendation tasks, and incorporate built-in bias mitigation for diverse, relevant, and fair recommendations. Understand the motivations behind using multi-task learning architectures and their benefits in improving model performance and task learning efficiency.