-
-
Our fully-connected web platform
-
All 5 machine learning models together in action
-
Unity project for interviewer sim
Inspiration
One of the biggest struggles people face is a fear of public speaking and interviews. In a world where so many aspects of our lives have been dominated, improved, and exacerbated by AI. But a field where we rarely see the use of AI is in fostering skills in socialization and real human connection. That is a fundamental issue we wanted to solve.
What it does
By signing up to the platform, any user regardless of their experience and confidence level can prepare for interviews and public speaking events with ease. For interviews, they have the option of providing specific details about the company they're working at and their role description, which the AI interviewer will ask useful, guided questions based on. The AI interviewer will track the user's speech patterns, facial expressions, eye contact, vocal discontinuities, and of course content in generating very useful, specific, and guided feedback as to how the user can improve in future interviews.
How we built it
We built a facial expression classifier model from scratch in PyTorch as well as an eye tracking model. We modified an existing lightweight architecture for facial detection and the GPT API for interview question generation and post-interview analysis. We output all of the generated responses and questions onto our sim character which is displayed to the user in a Skype call style.
Challenges we ran into
Integration! Combining all the different models was not an easy challenge and turning all the different data into text for the LLM.
Accomplishments that we're proud of
The amazing UI work and models we trained and combined.
What we learned
How to use different models to detect facial characteristics and emotions and connect multiple platforms to each other.
What's next for iNterview Pro
Helping the students today to become new hires tomorrow!
Log in or sign up for Devpost to join the conversation.