Guidance For Personalized Ecommerce Recommendations Using Amazon Bedrock Agents

Guidance for Personalized Ecommerce Recommendations Using Amazon Bedrock Agents

Table of Contents

1. Overview
   • Cost
2. Prerequisites
   • Operating System
3. Deployment Steps
4. Deployment Validation
5. Running the Guidance
6. Next Steps
7. Cleanup

Overview

The Sales Agent on Bedrock draws inspiration from the Rufus Sales Agent, which enhances the shopping experience on Amazon.com. Leveraging cutting-edge generative AI, this solution is designed to deliver personalized and dynamic interactions that drive product discovery and sales. By integrating Amazon Personalize, the Sales Agent provides tailored product recommendations, while Amazon OpenSearch Service enables rapid and accurate search capabilities across your catalog. Together, these technologies create a seamless and engaging customer journey, boosting conversions and satisfaction. The Sales Agent empowers businesses to harness Amazon’s proven AI expertise, transforming their apps and websites into highly effective sales channels.

The architecture of Sales Agent Rufus on Bedrock is illustrated below: 

Cost

Cost Considerations Breakdown When estimating costs, several key factors need to be considered:

1. Storage Costs

   • Amazon S3: Estimated storage of 10 TB.
   • DynamoDB: Stores 10 GB of product data and 1 GB of user data.
   • OpenSearch Serverless: Contains 10 GB of product data.
   • AWS Personalize: Estimated 1-hour training time to process 10 GB of data.

2. Model Training Costs

   • Assumes one training session per month, with each session taking 8 hours to complete.

3. Embedding Process Costs

   • Calls Titan model to generate embeddings.
   • Each request processes 100K tokens as input.
   • The embedding process consumes:
     • 100 OCU indexes in OpenSearch Ser