Hitachi AI Document Classifier

Using Gemini API, our project can take in and classify documents in a quick clean manner with built in HITL features to increase accuracy over time.

Comment

Inspiration

Based off project rules and a similar discord bot I made about 2 months ago.

What it does

Takes pdfs, pngs, or jpg/jpegs and checks their rating with googles Gemini to securely check document confidentiality.

How we built it

A backend with python that dynamically reviews documents to then prompt Gemini to accurately

Challenges we ran into

Bugs with frontend visuals and issues with incorrect prompting/API call loops

Accomplishments that we're proud of

A very accurate classifier!

What we learned

Utilizing APIs effectively, security, docker deployment.

What's next for Hitachi AI Document Classifier

Try it out!

Built With

Share this project:

Updates

posted an update —

In video Frontend API does not work for HITL system, but the backend does. Errors from some last minute frontend changes. It essentially takes the HITL responses and injects them into the base prompt for the Gemini API, depending on the specific dynamic prompt. This allows SEC to adjust the parameters of the model easily without needing to access the backend.

Log in or sign up for Devpost to join the conversation.