SpeakHire

Main page of application

Inspiration

Our inspiration for SpeakHire came from the common demand for a faster and more efficient hiring process. Usually, this process can be very time-consuming and impacted by human bias, so we thought of an AI-powered system that could perhaps automate a preliminary interview process, save time for recruiters, and even provide objective assessments of candidates.

What it does

SpeakHire assumes that a set of instructions was given by the employer, specifying questions and criteria through which candidates will be judged. Meanwhile, candidates are able to speak into their microphones to respond to the interview questions. The AI will then evaluate these spoken responses in real-time, using the company’s specifications to assess the quality of each answer.

How we built it

We built SpeakHire in distinct parts. The team worked collaboratively on multiple components of the project:

Backend: A Flask server handles data using Python.
Frontend: Using React, we gather audio data and convert it into a transcript.
Processing: The transcript is sent to the Flask server, where it's evaluated by the OpenAI API. The evaluation is then sent back to the React web app for display.
Styling: CSS and styling were added to enhance the user experience, ensuring the platform not only works smoothly but also looks appealing.

Together, React, Python, and CSS create the functional and visually engaging SpeakHire platform.

Challenges we ran into

One of the challenges we ran into was regarding backend/frontend integration. Aligning the diverse set of tools we used for backend into the frontend was complex. Another challenge we faced was with version control. As the project grew, we ended up making a couple of mistakes with how we were handling our repository.

Accomplishments that we're proud of

One of the things that we are especially proud of is that the AI evaluates spoken responses within seconds and gives the recruiters insights on how the candidate performed, potentially saving hours. Another thing we are proud of is that this system eliminates human bias through evaluating responses based on predefined criteria. On top of that, we are also proud that we were able to present this project with a user-friendly interface.

What we learned

We learned to deal with speech recognition challenges, such as background noise, through implementing a noise filtering system. We also learned how to work and collaborate in a team. Even though we were all doing our own parts, we were still able to communicate and integrate everything.

What's next for SpeakHire

We plan to integrate business-specific interfaces, allowing companies to add their own interview specifications and role criteria, further customizing the platform to meet diverse hiring needs.