Zero Inbox

• 

  zeroInbox cover

Inspiration

We received a lot of emails and it's tough to manage important ones while filtering out the crap. GenAi is able to read data and format it.

What it does

Thanks to GMail API, ZeroInbox fetches emails. Then feed them to Gemini to sum them up. Finally display the JSON sum up in a web interface

How we built it

• share knowledge and files and sync with the team on Google Chat
• work on the Google Ai console to try prompts and JSON-schema
• extract email datasets to train our models (Emron's email dataset)
• work with GMail API, Vertex API, Gemini API, colab and store resources on google Drive
• transform the JSON into html thanks to a templating tool
• display the final content thanks to web technologies providing a great UX

Challenges we ran into

• Gemini sometimes failed to respect instructions such as producing with the Json format given
• dev resources were not really available to play with Gemini in a simple manner

Accomplishments that we're proud of

• build something in less than 2 days
• bring value quite easily

What we learned

• Gemini fundamentals
• GenAI brings a lot of value to structure unformatted information

What's next for ZeroInbox

• we will pitch the product to VCs
• we will explore more use cases for data synthesis

Built With

• colab
• gemini
• gmail
• google-chat
• javascript
• json-schema
• material-design
• python
• vertex