MockAI

Inspiration

What it does

This is a platform where users can input their resume and a link to a job posting that they are interested in. Our AI interviewer agent reads the content from the job posting, using the candidate's resume as context to ask specific questions that the candidate is likely to be asked should they be selected for an interview.

How we built it

We used the GPT-4o model in order to analyze the job posting and resume to generate personalized questions for the role. We used Hume AI's EVI tool to handle the live conversation between interviewer and recruiter, covering emotional nuances to assess confidence and strength in their answers. All of this was bundled and hosted on Streamlit for an easy user experience.

Challenges we ran into

Some challenges that we ran into while building the project was handling various file types. The GPT-4o model can't handle pdf inputs directly for context. We didn't want to parse resumes for their text because of the varying number of styles, so we had to figure out a way to turn pdfs into images for seamless processing. We also had to figure out a way to be able to handle a live conversation on EVI with Streamlit.

Accomplishments that we're proud of

We're proud of our agent's ability to generate realistic questions on interviews. Our team inputted resumes and job postings that we were previously interviewed for. We found that on average, the agent was able to replicate 7 out of the 10 questions that we were actually asked in an interview, showcasing the realism that this product can emulate.

What we learned

We all improved heavily on our prompt engineering and front end engineering skills throughout this project. We learned how to use Streamlit in order to host our Python code, holding the logic for question and answer generation. We also learned how to combine multiple APIs to build out our features , namely the OpenAI API and the Hume API.

What's next for AceInterview

Built With

• hume
• openai
• python
• streamlit