Retail AI Insights: Demand Forecasting & Product Recommendations with AWS
This project builds a cloud-native, AI-powered retail intelligence system that predicts product demand and recommends items to users, all using serverless AWS services and scalable MLOps practices.
It is organized into 5 clear phases to demonstrate real-world infrastructure, AI integration, and automation.
Project Objective
Retail businesses lose revenue due to:
Overstocking or understocking inventory
Missed cross-selling opportunities
Manual forecasting processes that canโt scale
This solution uses AWS Bedrock, Amazon Personalize, and other cloud-native tools to:
Forecast product demand based on sales patterns
Recommend relevant products to users in real-time
Maintain a scalable, automated, and secure backend
Phase Description 01-data-ingestion Stream raw sales data using Kinesis Firehose into the S3 Bronze zone 02-data-preparation Clean and structure data using AWS Glue; output to Silver & Gold zones 03-demand-forecasting-with-bedrock Use Bedrock models on EC2 to generate daily inventory forecasts 04-product-recommendations-with-personalize Recommend items via API using Amazon Personalize 05-security-and-observability Ensuring IAM roles with least-privilege, running worklodas in private subnets, VPC endpoints, monitoring for the entire system
How to Deploy
Each phase directory contains its own setup guide using AWS CDK and scripts. Ensure:
You have AWS credentials configured
CDK is bootstrapped in your region
Python and Node.js dependencies are installed
Why This Project Matters
This project reflects the real-world needs of modern cloud systems:
Combines AI + Serverless for intelligent automation
Demonstrates MLOps best practices in a practical context
๐โโ๏ธ Contact
Created by Mohsin Sheikhani
From Code to Cloud | AWS Cloud Engineer | AWS Community Builder | Serverless & IaC | Systems Design | Event-Driven Designs | GenAI | Agentic AI | Bedrock Agents | 3x AWS Certified