Tecton

Feature stores can be essential components of your infrastructure for recommender systems. They simplify the process of deploying and managing RecSys models. But users often wonder which one is better suited for their use cases - Feast or Tecton? These two products are very different and address different requirements. In this workshop, we’ll give a hands-on, step-by-step example of building RecSys models using both Feast and Tecton. And we’ll provide our recommendations on when to use which product.

Join us in this panel discussion to hear from ML practitioners on their journey to implementing Recommender Systems. We’ll discuss the most common challenges encountered when getting started, and best practices to address them. We’ll explore organizational dynamics, recommended tools, and how to align business requirements with technical capabilities. You’ll hear about approaches to phasing in Recommender Systems, starting small and progressively iterating on more sophisticated solutions.

We’ll provide an introduction to Monolith, a system tailored for online training. Our design has been driven by observations of our application workloads and production environment that reflects a marked departure from other recommendations systems. Our contributions are manifold: first, we crafted a collisionless embedding table with optimizations such as expirable embeddings and frequency filtering to reduce its memory footprint; second, we provide an production-ready online training architecture with high fault-tolerance; finally, we proved that system reliability could be traded-off for real-time learning. Monolith has successfully landed in the BytePlus Recommend product.