FairFi - A tool to reduce bias in financial services

dashboard

Inspiration

Our inspiration stems from the alarming disparities in unemployment and poverty rates within African American communities. These inequities are deeply rooted in the racial bias of financial institutions. Studies show that African Americans are often subjected to discriminatory treatment when seeking financial assistance, such as loans, mortgages, and credit approvals. This bias results in fewer opportunities for wealth accumulation and financial stability, perpetuating a cycle of economic hardship. Our mission is to bring attention to and address the pervasive racial biases that hinder financial inclusion and economic justice for marginalized communities. We aim to empower companies to test and identify racial biases in their everyday operations, enabling them to address these disparities so that they may continue fulfilling their mission of serving and uplifting the communities they were created to support.

What it does

We have developed a website allowing companies to analyze their employees' biases by simulating a customer call. They may change the gender and accent of the AI-generated voice, and an AI-generated call will be made to the company using Twilio. The AI follows a script allowing pitch and tone to act as independent variables when the actual content is not. While generating the AI human-sounding voice, Twilio also transcribes the audio received from the employee into text.

We then used Cohere’s Command R+ to generate two outputs from the text. One is the sentiment of the employee and the other offers a general overview of the contents of the phone call. ...

How we built it

In the beginning, we brainstormed several ambitious ideas, but we ultimately chose FairFi because it struck the perfect balance between innovation and practicality, especially given our short timeframe. We started by establishing a clear workflow and designing the app architecture. For the front end, we utilized Next.js and Tailwind CSS. On the backend, we used Express and Node.js.

Challenges we ran into

Having to pivot our app because of our original plans being too ambitious. Using local tunneling to be able to get output from Twilio API.

Accomplishments that we're proud of

Our project integrates multiple APIs working seamlessly together, supporting advanced features such as real-time calling and sophisticated data analysis. We're especially proud of the complexity we’ve tackled and how much we’ve learned and grown throughout this journey.

What we learned

We gained valuable experience in integrating full-stack development with external libraries like Twilio and discovered the powerful capabilities of Cohere.

What's next for FairFi

Next we hope to add an authentication component so that only managers from a certain company may send calls to their own company. This allows us to avoid use for malicious intent. Additionally, we plan to display results in the form of graphs to allow easy representation of information.