Inspiration
In a world with an increasingly competitive job market, preparing for an interview can feel overwhelming, especially without someone to practice with. We created Cadence to give newcomers who are entering the job market for the first time a space to rehearse, practice, and improve at their own pace. Our goal is to help people communicate their experiences and career offerings clearly to help them feel prepared for their next potential opportunity.
What it does
-Personalized interview practice: Upload your resume and enter the job you are preparing for to generate your own unique interview scenario. -Natural conversation: Practice for real interviews with an AI interviewer who records and automatically transcribes your speech in real time. -Impactful feedback: Reviews user content, knowledge accuracy, and delivery, replaying the user’s strongest moments and areas to improve in.
How we built it
We first picked out a base 3D model and customized it to better fit the steampunk vibe. Then, we started building the basic structure of how our backend works: user sends in input like an answer to a question, the server calls Gemini API to answer/graded, continue. Once we continued training our API until we reached a point where we felt like the answers/grades/feedback all made sense. We then added on a computer vision feature that tracks your eye contact and body language to achieve a more accurate confidence rating. Finally, we focused on the animation of our 3D mascot, Gearwell and the UI/UX aesthetic of our website.
Challenges we ran into
Some of the biggest challenges of this project was finding a way to synchronize the video recording with speech transcription, and incorporating that input with AI to generate a natural conversation replicating a real interview. Testing also revealed that our AI would repeat questions and prompts that would be too heavily focused on resume details instead of the targeted role that the user was applying for. We also had to come up with a way to verify the accuracy of user input and find a way to rate the delivery of the user based on behaviors like their body language and eye contact.
Accomplishments that we're proud of
We’re proud to have built a functional interview practice platform that combines personalized questions, conversational AI, and local recording-based feedback. We implemented features like live-time transcription, continuous video recording, and clipping interview moments into one experience.
What we learned
We learned how to implement AI generated conversations, live-time speech transcription, and video/audio input into a dedicated user experience. User testing and debugging allowed us to train our API to connect user input with its generated feedback to improve the general interview experience.
What's next for CadenceAI
We would move to a higher tier of Gemini or switch to a different LLM. That way our grading, responses, and feedback can all be upgraded for a better user experience. We may also add more models to the website to improve accuracy.
Log in or sign up for Devpost to join the conversation.