dmnd-AI

Analyzing diamonds with AI

Comment

Inspiration

Even though it is a big part of Hong Kong’s retail industry, jewelry shopping can be a stressful experience for many people.

One of the problems many people face is that they are worried that they are making a bad purchase. For example, when a man is buying a diamond ring to propose to his girlfriend. He often knows nothing about diamonds. Yet, he has to spend a sizable portion of his salary to get one.

We want to revolutionize the jewelry shopping experience by using AI to empower shoppers with more transparency and giving them the peace of mind.

So we built dmnd-AI (pronounced Diamond AI), a web app that uses AI to help diamond buyers to make the right choice.

What It Does

dmnd-AI analyzes and compares the diamonds uploaded. It offers its recommendation and explains the reasonings. It also generates a list of actionable points for the shopper to look out for.

Here’s how it works:

  1. Upload an image of the GIA certificates. GIA certificates are the industry standard to show the authenticity of the diamonds.
  2. We extract and verify all the technical details on the certificates, including the diagram.
  3. A report with detailed analysis that includes proportion calculations, optical performance and other nuances will be generated.
  4. A recommendation is made using all the information.

How we built it

  • AWS Bedrock: We use Sonnet 3.7 for both the data extraction and analysis. We tried other models like Opus, however, Sonnet 3.7 gives the best accuracy and speed.
  • AWS Lambda: For the serverless functions.
  • AWS S3: File storage for certificates and reports.
  • AWS Amplify: Cloud deployment.
  • AWS CloudWatch: Observerability and debugging.
  • React, Tailwind CSS

How we used Kiro

  • Multi-language Translation: dmnd-AI supports English, Simplified Chinese and Traditional Chinese. We are able to do this in such time constraint is because Kiro makes doing translation really easy.
  • UI Design: Using Kiro, we are able to quickly prototype an UI without using any design tools. We simply used screenshots and words and it was able to generate production ready UI.
  • Version Control: Instead of using actual Git commands, we are able to do version control via natural language in Kiro.
  • Refactoring: Kiro is able to assess the quality of our code. Pointing out the strengths and weaknesses. It also listed high priority issues, optimization and security suggestions.

How we used Amazon Q

  • Brainstorming: We used Amazon Q to design the architecture of dmnd-AI. It provided a list of suggestions such as AWS infrastructure services that would be helpful in our project.
  • Setup: Amazon Q helped us setup AWS services like S3 and Lambda.
  • Debugging: Using AWS CLI, Amazon Q was very useful in debugging backend issues.

Built With

Share this project:

Updates