LogicCart Website Change Request Hub

AI-powered Website Change Request Portal for E-commerce Teams

Comment

LogicCart Website Change Request Hub

Inspiration

This project was inspired by a real-world pain point observed in cross-functional e-commerce teams:
Requesting website changes via email often leads to confusion, delays, and inconsistent formats.

As a UX/UI designer, I’ve experienced how small miscommunications can derail timelines, especially when developers lack clear, structured input. I wanted to solve this with AI-powered clarity and a streamlined workflow.

What I Built

LogicCart is a secure, serverless platform that allows team members to submit website change requests in a structured form, receive real-time AI feedback, and view request statuses instantly on a dashboard.

Key innovations include:

  • AI Preview Analysis powered by Amazon Bedrock Nova Lite, providing policy compliance checks, missing info detection, and acceptance criteria generation.
  • Post-submission AI decision making powered by Amazon Bedrock Nova Lite, integrated with SNS for automated approval or clarification emails.
  • A serverless architecture with 6 AWS Lambda functions, API Gateway, DynamoDB, and S3 buckets for secure image uploads and policy documentation.
  • A real-time dashboard that pulls live request data, enables CSV/PDF export
  • Secure user login using AWS Cognito, with session validation via CloudFront Functions and PKCE flow.

What I Learned

This was my first end-to-end AI project using Amazon Bedrock + full AWS infrastructure, and I learned:

  • How to design and invoke Bedrock Runtime API via Lambda for real-time decision making
  • The importance of clean IAM permissions and security measures
  • With AWS Kiro, I managed to execute secure S3 upload via presigned URLs, API Gateway, Cloudfront and Cognito authentication integration.
  • Real-world prompt tuning and UX optimization using Kiro to co-build Lambda logic, UI flows, and Bedrock prompts

Challenges I Faced

  • Security setup: Configuring CloudFront, Cognito, and session validation was challenging.
  • Data integration: Mapping user-submitted data, image metadata, and AI analysis results to a unified schema in DynamoDB.
  • UX polish under time pressure: Ensuring responsive design, mobile-first layouts, and accessible UI while building a functional full-stack system.

Final Reflection

This project brought together UX/UI, AI, and serverless engineering into a working prototype that could be adopted by real teams. Watching the AI suggest improvements before mistakes happen, that’s real impact.

I hope this project demonstrates the power of Amazon Bedrock + Kiro for rapid, real-world AI agent development.

Built With

Share this project:

Updates

Submission history