ResumeRound

• 

  Landing Page

• 

  Performance Audit

• 

  AI Interviewer

Inspiration

AI is evolving rapidly and so is job competition. Hundreds of applicants now compete for a single role, and the bar keeps rising. As a vibe coder — someone who's genuinely good at building full-stack products but not always the strongest at articulating it in a high-pressure interview — I felt this personally. Generic mock interview tools ask the same scripted questions regardless of who you are or what you've built. That gap is what ResumeRound was built to fix.

What it does

ResumeRound conducts a real-time, voice-native mock interview personalised entirely to you. Upload your resume and paste the job description you're targeting — the AI reads both before asking a single question. Every question references something you actually claimed. It follows up when answers are shallow, probes your specific experience, and at the end gives you a full performance audit: Verified Strengths (claims you defended well), Probe Zones (where a real interviewer would push harder), and a readiness summary. No typing. Just talking.

How we built it

Built with React 19, Tailwind CSS 4, and the Google Gemini API. The voice interview runs on gemini-2.5-flash-native-audio-preview for real-time zero-latency audio. PDF parsing and post-interview report generation runs on gemini-flash. The entire flow — landing page, resume upload, live voice session, performance report — is a single seamless web app.

Challenges we ran into

Getting truly low-latency voice interaction that feels like a real conversation, not a chatbot with pauses, was the core technical challenge. PDF parsing that extracts specific claims accurately enough for the AI to reference them meaningfully in questions also required careful prompting and context handling.

Accomplishments that we're proud of

The "Probe Zones" feature — the AI doesn't just tell you what you did well, it tells you exactly which claims on your resume a real interviewer would challenge and why. That specificity is what separates ResumeRound from anything else out there.

What we learned

Voice-native AI experiences require a completely different design approach than text interfaces. Silence, pacing, and natural follow-up matter as much as the questions themselves. Prompting for depth-based dynamic questioning — where the AI follows up based on your actual answer — is a genuinely hard problem worth solving well.

What's next for ResumeRound

Multi-round interview modes (behavioural, technical, culture fit), company-specific interview style simulation, and eventually a B2B layer where companies use the same engine to screen candidates at scale.

Built With

• framer-motion
• gemini-2.5-flash-native-audio
• google-gemini-api
• lucide
• react
• tailwind-css
• typescript
• vite