Today, Amazon Personalize announces the general availability of two new recipes, User-Personalization-v2 and Personalized-Ranking-v2 (v2 recipes). Built on Transformers architecture, these new recipes support catalogs with up to 5 million items with lower inference latency. Amazon Personalize testing showed that v2 recipes improved recommendation accuracy by up to 9% and recommendation coverage by up to 1.8x compared to previous versions. A higher coverage means Amazon Personalize recommends more of your catalog. These new recipes also support item metadata like genres and descriptions in inference responses, allowing customers to easily enrich recommendations in their user interfaces.

Amazon Personalize enables customers to personalize their website, app, emails, and more, using the same machine learning (ML) technology used by Amazon, without requiring any ML expertise. Using recipes - algorithms for specific uses cases - provided by Amazon Personalize, customers can deliver a wide array of personalization, including product or content recommendations and personalized ranking.

To get started with Amazon Personalize, provide an activity stream – clicks, page views, signups, purchases, and so forth – as well as a catalog of the items you want to recommend, such as videos, products, articles, or music. You can also provide demographic information from your users. Amazon Personalize will process the data, train and optimize your custom model, then host it for your applications.

User-Personalization-v2 and Personalized-Ranking-v2 are available in all supported regions. You can access them through Amazon Personalize console or API. To get started, refer to our documentation.