Skip to content Skip to footer

Amazon Personalize introduces new features that support larger product catalogs and decrease latency.

Amazon Personalize, a machine learning (ML) technology used for customizing user experiences, has announced the general availability of two advanced recipes named User-Personalization-v2 and Personalized-Ranking-v2. These features utilize the Transformers architecture to support larger quantities of item catalogs with lower latency.

The new features are improvements on previous versions, particularly in terms of scalability, latency, model performance, and functionality. They are capable of handling training with up to 5 million item catalogs and 3 billion interactions, making them suitable for large-scale operations. They are also optimized to reduce inference latency and expedite the training of large datasets. Amazon has tested these features and seen a 9% improvement in recommendation accuracy and 1.8x improvement in recommendation coverage.

The new recipes automatically include item metadata in inference responses, enabling more contextual recommendations. The metadata includes information such as the description and availability of items, as well as the genre of content. This allows large language models to understand product attributes and make more relevant recommendations.

The process of utilizing these features involves setting up Amazon Personalize resources and creating a dataset. Upon doing this, one can then select the desired recipe, apply any necessary configurations and begin training. Users can make a selection from a list of recommended items, which are arranged based on a user’s likelihood to engage with them. Following use, users should ensure they clean up any unused resources to avoid clutter and disarray in their workspace.

This upgrade to Amazon Personalize increases support for larger item catalogs, reduces latency, and optimizes performance. This advanced solution uses ML technology and employs customized models, aiming to create a satisfying user experience by presenting highly relevant content recommendations.

The team behind this enhancement includes Jingwen Hu, Daniel Foley, and Pranesh Anubhav among others. These professionals specialize in developing ML systems, recommender systems, and AI services. Their expertise in their respective fields has contributed to the advancement and better performance of the Amazon Personalize feature. They aspire to continuously improve this technology to meet users’ evolving needs.

Leave a comment

0.0/5