SkillNova Voice: Real-Time AI Interviewer

Practice technical interviews with an AI that talks back. Powered by Gemini 1.5 & ElevenLabs for real-time voice interaction and instant "Hiring Decision" feedback.

Comment

💡 Inspiration

We've all been there: The "Technical Interview." It’s not just about knowing the code; it’s about explaining your thought process under pressure. Most prep platforms only check your syntax. I wanted to build something that checks your presence.

I built SkillNova Voice to simulate the actual feeling of a Zoom interview with a Senior Engineer—awkward pauses, follow-up questions, and immediate feedback.

🤖 What it does

SkillNova Voice is a real-time, voice-first technical interviewer.

  1. It Speaks: You don't type. You talk. The AI uses ElevenLabs Conversational AI to speak to you with human-like intonation.
  2. It Thinks: Powered by Google Gemini 1.5 Flash, it doesn't just read a script. It listens to your answer, understands the context, and asks dynamic follow-up questions.
  3. It Decides: At the end of the session, the AI generates a "Hiring Decision Report" that analyzes your technical accuracy, communication clarity, and gives a final Hire / No Hire verdict.

⚙️ How we built it

The app is a high-performance React frontend hosted on Netlify.

  • The Brain: We used Gemini 1.5 Flash for its low latency and massive context window. It acts as the "Interviewer Persona," maintaining the state of the conversation and evaluating answers against a hidden rubric.
  • The Voice: We integrated ElevenLabs via WebSockets to stream audio bi-directionally. This allows for near-instant response times, making the conversation feel natural rather than robotic.
  • The UI: Built with Vite and Tailwind CSS, featuring a glassmorphism design that keeps the focus on the voice interaction.

🧩 Challenges we ran into

The biggest challenge was Deployment & Latency.

  • State Management: coordinating the "Listening" state between the browser's microphone API and the ElevenLabs stream was tricky.
  • Deployment Purgatory: We faced significant issues with environment variable injection on Netlify (Vite handling vs. Server-side handling), which taught us a lot about secure key management in frontend builds.
  • Prompt Engineering: getting Gemini to be "strict but fair" took several iterations. Initially, it was too nice! We had to tune the system prompt to act like a real, skeptical Senior Dev.

🚀 What's next for SkillNova Voice

We plan to add specific "tracks" (System Design, Python, Java) and integrate a resume parser so the AI can ask questions based on your specific background.

Built With

Share this project:

Updates