Retailizy : Agentic AI based Retail Catalog Builder
Catalog Building Tasks
Retailizy - Makes Catalog Building Easy
Retailizy- High Level Architecture

Inspiration

Small retail store owners faces challenges to build their product catalog . Retailers spend weeks manually writing catchy taglines, product descriptions, guessing target audiences, and planning seasonal campaigns. What if AI could do all of that in seconds — right inside the catalog management tool?

What it does

Retailizy is an AI-Agents powered retail catalog builder that turns a basic product entry into a complete, market-ready listing. Built on Amazon Bedrock with Nova Lite and Nova Canvas, it helps retail teams: User Persona Diagram

Features

Feature	Description
AI Tagline Generator	Generates catchy one-liner marketing taglines for products
Category Suggestion	Suggests best-fit product categories from existing catalog
Demographics Analysis	Identifies 2–4 target customer demographic profiles per product
Cross-Sell Recommendations	Suggests related products for cross-selling
One-Click Catalog Entry	Runs all 4 AI features in a single agentic pass
Sale Event Suggestions	Recommends promotional events for a category based on demographics
Review Sentiment Analysis	Analyzes customer reviews and extracts improvement insights
Improvement Suggestions	Generates actionable product improvement recommendations from negative reviews
JIRA Ticket Creation	Auto-creates JIRA tasks from AI improvement suggestions
Campaign Calendar	Generates a 3-month AI campaign plan per product
Promotional Email Draft	Writes a promotional email for a campaign event
Campaign Image Generation	Creates campaign ad images using Amazon Nova Canvas with text

How we built it

Retailizi has used AI agents created using Strands library for all AI features in the application. agents have used AWS stack for AI models:

Amazon Bedrock Platform for access Nova models
Amzon Nova Lite to build agents for Catalog, Campaign, CRM services
Amazon Nova Canvas for text to Image feature
Amazon Dynamo DB for database
Amazon S3 for file and image storage
Strand to build agents
Fast API for Backend API services
NextJS for frontend application

Agent Responsibilities

CatalogAgent

Method	What it does
`generate_tagline`	Writes a marketing tagline for a product
`suggest_category`	Picks the best category from the existing catalog
`find_demographics`	Identifies target audience (age, profession, lifestyle)
`suggest_related`	Recommends cross-sell products from the catalog
`suggest_sale_events`	Proposes seasonal promotions for a category
`build_full_catalog_entry`	Runs all 4 product tasks in sequence (one-click enrichment)
`build_category_profile`	Combines demographics + sale events for a whole category

ReviewAgent

Method	What it does
`generate_reviews`	Generates realistic customer reviews for a product
`generate_reviews_batch`	Generates reviews for all products with none
`improve_review`	Analyses a negative review and suggests product improvements
`create_jira_ticket`	Creates a Jira issue from an improvement suggestion

ImageAgent

Method	What it does
`generate_campaign_image`	Generates a campaign ad image via Nova Canvas, saves to S3

Challenges we ran into

Amazon Nova site ( https://nova.amazon.com/) is not available in India region. Therefore I used Amazon Bedrock platform to access Nova Models.

Accomplishments that we're proud of

Seamless integration of Strands framework with Amazon Bedrock . I created multiple features in an easy to use web application covering multiple aspects of Catalog building