InterviewHero

An AI-powered interview coach that turns any job link into a live practice interview and gives instant STAR feedback to help students grow confidence and land opportunities.

Comment

Interview Hero

Our Project Story

We wanted to solve a problem every student faces: interview anxiety and the lack of personalized, affordable interview preparation. Campus career centers are amazing, but they can’t support everyone 1:1, especially during internship season.

We asked ourselves:
“What if any student could get a personalized mock interview in seconds, anytime, for any job?”

That question inspired us to build an AI-powered real-time interview coach.

How we built it

  1. Job Link Parsing

We built a scraper that extracts role responsibilities, required skills, and qualifications directly from a job posting. This data becomes the foundation for dynamic interview generation.

  1. Real-Time AI Interviewing

We used LiveKit for low-latency audio streaming, enabling natural two-way conversation. Combined with Gemini, the interviewer adapts to:

your answers

missing details

behavioral cues

requested clarifications

This creates a fully responsive mock interview experience.

  1. AI Transcript Evaluation

After the interview, our system generates:

STAR-based feedback

Communication clarity scores

Confidence scoring

Personalized next steps

Key strengths + areas to improve

📚 What We Learned

How to build real-time audio AI agents using LiveKit

How to orchestrate AI models for dynamic interviews

How to convert raw transcripts into structured, actionable insights

How crucial latency, context windows, and prompt engineering are

How to design tools specifically for student success and confidence-building

And most importantly: We learned that good AI isn’t about replacing humans — it’s about scaling access to support that students traditionally don’t have.

Challenges We Faced

  1. Real-Time Latency

Maintaining low audio latency was difficult. Even a small delay disrupted conversation flow.

  1. Parsing Inconsistent Job Descriptions

Job postings vary wildly in clarity and structure — we had to build logic to standardize them.

  1. Adaptive Questioning

Teaching the AI to ask follow-up questions based on what the user actually said required careful prompt design and fast model response times.

  1. Reliable Transcript Scoring

We had to ensure the feedback was consistent and didn’t contradict itself, which required multiple scoring passes.

  1. Time Constraints

Integrating scraping, real-time audio, and post-interview evaluation in a single weekend… not easy, but incredibly rewarding.

Final Thought

We built this project to make interview prep accessible, personalized, and sustainable for every student. Our hope is that it becomes a tool that boosts confidence, unlocks opportunities, and supports long-term student success.

Built With

Share this project:

Updates