• 

  Resona Interface

Resona — Trust-Aware AI Customer Support

Inspiration

Resona was inspired by the need for safer and more reliable AI-powered customer support. A chatbot should not simply accept a claim and promise a refund. It should verify the evidence, check the order, and escalate uncertain cases to human reviewers. This project explores how AI can support customer service while keeping important actions accountable.

What it does

Resona helps customers check order information and submit damaged-item claims using photo evidence. It evaluates image quality, compares submitted photos with product reference images, and uses policy-based decisions to support approval, photo retakes, or human review. It also tracks claim evidence and case events to improve transparency.

How we built it

Resona combines an AWS-based backend with a web chat interface. AWS Lambda handles API requests and business logic, Amazon API Gateway exposes the endpoints, Amazon DynamoDB stores order and claim information, and Amazon S3 stores photo evidence. Python and OpenCV power image decoding, quality checks, and visual feature comparison. JavaScript manages the customer-facing conversation and claim workflow.

Challenges we faced

One major challenge was connecting the chat interface, AWS services, and image-processing pipeline into a consistent workflow. We also had to handle blurry or dark images, invalid uploads, missing orders, repeated evidence submissions, and uncertain image comparisons. Another challenge was ensuring that the system does not claim a refund has succeeded unless the result is verified.

What we learned

Building Resona taught us how to integrate serverless AWS services, design API-driven workflows, process images with OpenCV, and manage application state across customer interactions. We also learned the importance of clear error handling, evidence tracking, and human escalation when automated decisions are uncertain.

Future improvements

Future work includes improving product-matching accuracy, completing end-to-end order and refund verification, strengthening the human-review workflow, and expanding automated testing and deployment checks.

Built With

• ai
• amazon-api-gateway
• amazon-dynamodb
• amazon-web-services
• amazon-web-services-(aws)
• aws-lambda
• computer-vision
• css
• customer-support-automation
• html
• human-in-the-loop
• image-processing
• javascript
• numpy
• opencv
• python
• rest
• serverless-architecture
• trust-scoring