CallSafe

Inspiration

Last summer, Will received a call from his grandmother in distress. She had just been called by a man claiming to be her grandson, and that he needed $5000 to be bailed out of jail. When she asked if it was Will calling, the impersonator confirmed that it was. The caller emphasized that he needed the money as soon as possible, and to not tell Will's parents. Luckily, Will's grandmother knew something wasn't right, and hung up the phone to call him instead. Sadly, many of these situations don't have a happy ending; in 2022 alone, Americans lost 40 billion dollars to scam phone calls. Our goal is to provide a solution to this issue, as there is no reason for this problem to be so prevalent in the age of artificial intelligence.

What it does

CallSafe uses artificial intelligence to detect and deter scam phone calls.

How we built it

We made a mobile app with Flutter and Dart, a database with Firebase, and a backend and AI Model with Python and Flask

Challenges we ran into

There is very little data available for scam call transcripts, thus we had to generate our own to supplement what we could find online. We also had issues sending files over our self-made API and had to reformat the code base to provide this functionality.

Accomplishments that we're proud of

Creating our own dataset of over 1000 scam call examples and 3000 legitimate phone calls. Clean-looking UI with full functionalities.

What we learned

We learned how to fine-tune transformer models by generating and tokenizing a dataset. We also learned how to make Python and Flask work with Flutter, a HuggingFace AI model, Firebase and an API.