The problem I'm solving is that moving customer information into new systems when setting up accounts is usually slow and full of mistakes, which frustrates everyone involved.

I happened to have a research project on general data transformation. So I built a quick demo to tackle this issue.

The demo uses Claude 3.5 and a prompt framework to transform customer CSV files into your application's data model. The UI is built with Streamlit.

I started with CRM data as an example, but I'd love to expand to more complex use cases with various data types. If you have interesting scenarios, let me know.

Here's the research behind it: https://cocoon-data-transformation.github.io/page/transform The project is open sourced: https://github.com/Cocoon-Data-Transformation/cocoon