Inspiration
I was talking to my girlfriend about her experiences going out. She said she carried pepper spray, but felt it might not be a complete solution. I thought it would be helpful if it could contact the police somehow and looked into it. We ended up finding pepper spray and jewelry that would signal an agent to call emergency services, the catch being a monthly subscription fee. Considering the device may never even be used, paying a monthly fee felt pointless. Sid and I realized an AI agent, engineered with safety in mind, could easily handle such a task. Intelliwear removes the need for a subscription fee, and helps millions across the world feel safer at a reasonable price.
What it does
An app contains a survey to get the user’s emergency information as well as a trigger to make the call, in place of a Bluetooth connected button on pepper spray or jewelry. When the call is triggered, a message with the user’s information is passed to the agent handling responses. This allows the agent to correctly handle each user.
How we built it
IntelliAlert has two components: it's front end is a mobile application written in Swift for iPhones. The backend is built in Python using FastAPI and the Twilio Voice SDK.
Twilio allows us to place outbound calls from an AI agent. This agent makes a call to an emergency responder (Siddharth for the purposes of this hackathon), and relays information about why the panic button was clicked.
Then, we use speech-to-text capabilities and Groq to craft appropriate responses given the users information.
Challenges we ran into
This project is the first time either of us have participated in a hackathon, so the idea of building a project in limited time was a major hurdle for us. Also, neither one of us had experience working with Swift for mobile app development, so we had a lot of learning to do. At the tail end of the hackathon, Twilio unexpectedly stopped working for us, so we weren't able to show off the full capabilities of our app.
Accomplishments that we're proud of
We were able to get the Swift app working which was a nice win considering it was our first time using Swift. While we are unable to prove it, getting the agent to respond to our question over the phone call was our greatest accomplishment of the night.
What we learned
We learned a fair amount of Swift, and reviewed EC2, Route 53, and API setup.
What's next for IntelliAlert
We would like to work on adding context to our agent by streaming the phone’s microphone data to our backend. We would also like to prototype the physical device and add capability, similar to the Tesla phone key, so there is a constant connection after the initial handshake.
Log in or sign up for Devpost to join the conversation.