Inspiration

Students and young professionals often feel unprepared for interviews—not because they lack ability, but because practice is intimidating. Career coaches can be expensive, mock interviews feel awkward, and many people are afraid of being judged. We wanted to create a safe, realistic way to practice interviews anytime, anywhere.

What it does

Users select a role and company, which the backend matches to a stored job description. They then upload their resume, which is parsed for key information. All of this forms the context for the AI, which generates relevant, realistic interview questions. Users can then initiate a real-time voice call with the agent, and receive a detailed evaluation afterward.

How we built it

Resume Analysis Agent parses resumes to extract skills, experience, and projects. Question Prep Agent generates context-specific, role-relevant interview questions using the parsed resume and stored job descriptions.

Interview Agent uses the OpenAI Realtime SDK to conduct live voice interviews, maintaining a persistent connection for better context control and natural conversation flow.

All agents have access to tools including databases, NLP extraction, and question generation. The backend is built with FastAPI, and the frontend uses React and Redux for a responsive, real-time user experience.

Challenges we ran into

Trying to get realtime voice ai to work. Initially we explored agora ai, but later pivoted to gpt-realtime. Handling background noise and interruptions during real-time voice conversations Generating questions that are specific to the resume, not generic The lack of sleep. We slept like 2 hours.

Accomplishments that we’re proud of

Building a fully real-time, voice-based interview experience. Generating personalized interview questions directly from resumes. Finishing this 24 hour in person hackathon.

What we learned

We learned that realism matters more than perfection—users value natural conversation and relevant questions over flashy features. We also learned that feedback needs to be specific and constructive to be truly helpful, especially for early-career candidates.

What’s next for Interview Pal

Interview evaluation by AI agent, to give feedback on how the user performed in the interview. Support for different interview styles (behavioral, technical, case) Company-specific interview formats and difficulty levels Progress tracking across multiple interviews Multilingual support and accessibility improvements

Built With

Share this project:

Updates