Cadence

An AI powered interview helper that evaluates the quality of your answer as well as your interviewing skills themself. These include your body language, eye movement, and general speech confidence.

Comment 1

Inspiration

The most nerve wracking portion of the job application process is always the interview. The constant fear of messing up or not knowing the answer is a consistent worry among all job applicants. This worry prompts people to often fidget, break eye contact, or display unprofessional body language. To help prepare further and alleviate stress, we built Cadence.

What it does

Cadence is an AI-powered interview practice platform that helps users improve both their technical responses and communication confidence. It allows you to pick from a plethora of premade questions across various fields or input your custom questions. From this, it sets up a mock interview in which you answer the question, and it judges both your quality of your answer and your confidence score, giving you areas of feedback for both.

How we built it

We built the back and front end using next.js and tailwind CSS. We uploaded the audio to Gemini 2.5 Flash to transcribe it into text, labeling each filler word and large pause. Gemini then judges the accuracy of the answer to the question and gives feedback. The actual webm video file is sent to Twelve Labs Marengo model for video analysis, and it's given a score based on confidence and nervous tendencies by tracking movement.

Challenges we ran into

We often had many issues relating to the confidence score, specifically with the feature not counting filler words and their effect on the score. In order to fix this, we combined both the Gemini analysis and the Marengo video and adjusted our own formula to give a more accurate confidence score. Additionally, API keys and general API request form took a long time to get right.

Accomplishments that we're proud of

Our team is very proud of our use of Twelve Labs as it is a new tool to us, and we are very satisfied with how we implemented it. We believe our product is a genuinely helpful tool that helps interviewees learn better interviewing habits.

What we learned

We learned how to implement new tools with less public users, so we had to go more in depth into the documentation to figure out our issues. We strengthened our team work by assigning tasks to everyone for maximum efficiency, leading to a better end product.

What's next for Cadence

We would to make an account system that could log your sessions and keep track of how much prep time you have done. To make it more scalable, we also could allow for the input of multiple questions and have it do a multi-question interview. Overall we believe our product is very beneficial, but we could scale it even further and increase efficiency.

Built With

Share this project:

Updates