Scammer Jammer

Scammer Jammer is our solution to deter scam callers. Using Hume.ai's empathic voice interface, our app mimics the speech of a real human to annoy and waste scammers' time during real-time calls!

Comment

Inspiration

Our team loves creating interesting projects that defy the norm, and just like Hume.ai's innovative empathetic LLM, we wanted to create something that can leverage its unique power in a context that is both silly but also helpful to society!

What it does

The idea of the website is to create a solution to the vast number of scam calls that occur within the US. In our case, we have created a website where you initiate a call using a custom phone number provided through Twilio which will initiate the call between you (the scammer) and Hume.ai. Acting as the scammer you can try to convince the EVI to buy you things, but will encounter them as incredibly annoying. This is the point of the product, we wanted to create an incredibly frustrating person to talk to the scammer. We made it so that it would always talk about topics unrelated to the question and overall just waste the other person's time, so that the scammer would have less time to scam more vulnerable victims.

How we built it

We first began by understanding how Hume.ai works by going to their workshops. Afterwards, we began splitting up the work between front-end and back-end. We took care of the front-end first in Vite and Vercel, using it to guide our product vision and paying attention to the details and features we would want to include. The second half of the hackathon was more focused towards developing the backend, where we leveraged Node.js and Typescript's WebSocket interface to communicate audio data across secure channels.

Challenges we ran into

The biggest hurdle was to create a custom middleware server that would receive the input from Twilio's audio, transcribe the audio in a format that EVI could respond to, stream EVI's response to our middleware, and then stream the response as back to Twilio. This challenge was very hard, but luckily we had an amazing experience working with the Hume AI team where we received consistent and helpful advice, and we additionally solved some of the technical challenges on our own! At the end, we designed a proxy between Twilio and Hume.ai that could forward audio between the two clients while also displaying the conversation on our custom UI.

Accomplishments that we're proud of

Our biggest accomplishments are our well designed and clean front end, and our backend that has the ability to monitor the data and enable bidirectional communication / monitoring between Hume.ai and Twilio.

What we learned

Our group deepened our core understanding of API calls, especially in the context of Hume.ai's eLLM model and Twilio API, advanced our knowledge in topics such as proxies and WebSockets, and also expanded our technical framework knowledge such as learning Vite! Another major takeaway our team learned was to value the importance of any technical challenge as seemingly simple ideas are potentially difficult. Initially, our team thought the project idea was relatively straightforward, but eventually, we discovered the true engineering and significant thought process required to tackle our problem. Our group eventually created and designed a scalable architecture that could efficiently transport audio data in multiple forms and still provide low latency so that the conversation would feel natural.

What's next for Scammer Jammer

We are excited to continue Scammer Jammer by creating an app that can be both offered on iOS or Android devices! Our team strongly believes in the spirit of helping people, and wants to improve our application's effectiveness as a mobile app could intercept real spam and scammers. Our team plans to implement the app using similar techniques as our current website.

Built With

Share this project:

Updates