An agentic personalisation engine for commerce that learns why a shopper buys, not just what. We built it as a team of four at Cognizant's Technoverse 2.0, where it made the finals: top 24 of 5,600+ teams.

The browser streams consented events; agents on AWS Bedrock turn them into facts about the shopper; Titan embeddings go into OpenSearch Serverless; and a recommender ranks those facts, drafts an offer and checks every claim against its sources before it shows a single product.

What I built: the storefront (React, TypeScript), the event SDK that loses nothing to a reload or a closed tab, consent checked on every event in the browser and again on the server, a noise gate that keeps trivia away from the model, the move to OpenSearch Serverless, and a relevance cutoff so a search for "towel" stopped returning 48 shirts.

Write-up: https://rajveers.com/projects/hyperpersona

Code: https://github.com/rajveeerr/Hyperpersona

Live: https://hyperpersona-web.vercel.app