mail.AI

Inspiration

We were inspired by exceedingly long and frequent university emails to build a service that would allow us to easily summarize, extract information, and respond to emails.

What it does

mail.AI uses state of the art AI and Natural Language Processing models to generate a summary of emails, a response to an email, and to extract main themes or contact information from text. The response page can also be used to have a full conversation with a true artificial intelligence.

How we built it

We started off by building our main landing page and hosting it with Google Cloud. We decided to use Google Cloud Run to allow easy scaling up/down during periods of high/low activity. We also wanted to build a fully automatic CI/CD pipeline. We used docker images and BASH scripts to set our Cloud Run instance up such that it automatically builds a new Docker image from the Git Repo and publishes the image to the server whenever changes are bushed in the main branch in our Git Repo. Once we got that working, we built the rest of the pages of our website and build the integration with the OpenAI REST API. Our code sends PUT requests to the API based on the user input then filters the model output and sends it back to the user.

Challenges we ran into

We had significant difficulty setting up our Google Cloud Run instances. Google Cloud is not user-friendly at all, and none of us had used it in the past. We also had issues with getting the NodeJs OpenAI library to work, so we had to build our own system that builds REST API requests and sends them to the OpenAI server.

Accomplishments that we're proud of

We are very proud of the fully automatic CI/CD pipeline we built for our code that will allow us to easily maintain and improve upon it in the future. We are also very proud of how well the AI model works. The results are very intelligent and far surpassed our expectations. Lastly, we were very proud of the animations we integrated in our service.

What we learned

We learned a lot about Cloud Server Hosting. Namely, that we will never be using Google Cloud again. We also learned a lot about CI/CD pipelines and web development.

What's next for mail.AI

We plan to expand our web application with more pages and functionality, in addition to optimizing the web app to work on mobile browsers.

Built With

• bash
• css
• docker
• git
• google-cloud-run
• html
• javascript
• ml
• natural-language-processing
• node.js
• openapi
• react
• rest
• yaml