HARP | Devpost

Inspiration

The current solutions for automated technical interviews were lackluster. We wanted an all inclusive platform that combines an interactive code editor and runtime environment with an AI generated interviewer persona that would be conducting the interview.

What it does

HARP is an online platform where one can practice their skills during a mock technical interview while supervised by an AI persona. The AI persona can monitor cues such as typing behavior and facial expressions to provide a post-interview debrief and highlight areas for improvement.

How we built it

HARP is composed from a suite of microservices to provide its functionality. The microservices include the web platform, the AI persona service, the emotion recognition service, and the code development and execution service. All microservices were implemented using Node.JS and TypeScript and leverage existing open-source technologies such as Express, Next.JS and Tensorflow to accelerate their development.

Challenges we ran into

A challenge we ran into was incorporating the AI features into the platform. Namely, we wanted our AI persona to have a face and react in realtime to the interviewee’s behavior. However, most LLMs are based on messaging model, which made it difficult to make it feel like it was present in the moment with the user.

Accomplishments that we're proud of

We were successfully able to use Gemini to generate coding challenges to present to the interviewee. The interviewee would then be able to be able to type code into the editor and submit it for testing. We are also able to give a voice to our AI interviewer, and the AI interviewer is also able to parse and understand the facial expressions of the interviewee as they work on a solution.