Intervoice

• 

  Data Flow Diagram

Inspiration

Most job seekers spend hours preparing for interviews, but 70% feel generic practice tools leave them unprepared for questions about their real experience. Nearly three out of four candidates crave an app that helps them practice through real-time, voice-based mock interviews in an interactive way

Who It's For

Intervoice is designed for job seekers and career professionals who want to: Students & Graduates: Practice interview skills before entering the job market Career Changers: Prepare for interviews in new industries or roles Job Seekers: Get personalized interview practice for specific positions Professionals: Improve interview performance and confidence Recruiters & HR: Use as a training tool for interview preparation

what we built

Intervoice is an AI-powered interview preparation platform that uses 6 specialized AI agents to help you ace job interviews. It analyzes your resume, researches company-specific questions, generates personalized interview questions and STAR-method answers, then conducts real-time voice or text mock interviews using advanced AI. The platform provides detailed performance feedback and improvement suggestions, simulating the complete interview experience from preparation to evaluation.

How we built it

AI Agent Architecture: Implemented 6 specialized AI agents using Claude API and Gemini Live for real-time voice interactions

Backend Infrastructure: Built a FastAPI server , integrated Firestore for data persistence, and implemented WebSocket connections for real-time communication between frontend and AI agents.

Frontend Development: Created a Flutter Web application and WebSocket integration for live chat functionality

Voice Processing: Implemented MediaRecorder API for audio capture, base64 encoding for audio transmission, and integrated Gemini Live API for natural voice conversations with real-time streaming.

Data Flow Management: Developed a comprehensive workflow system that processes resume uploads through multiple AI agents, stores results in Firestore, and provides real-time feedback to users.

🤖 How We Used Claude

We used Claude 3.5 across 6 specialized AI agents:

Accomplishments that we're proud of

Creating a truly intelligent interview preparation system that uses multiple AI agents that feels like talking to a real interview coach.

Implementing intelligent error handling that handles API failures, network issues, and browser compatibility problems while maintaining a professional user experience.

Developing a comprehensive feedback system that provides detailed performance analysis and actionable improvement suggestions, making each practice session genuinely valuable

What we learned

AI Agent Orchestration: Building systems that leverage multiple specialized AI agents working together taught me how to design complex workflows where each agent has a specific role

Modular Testing Strategy: importance of testing individual components separately - testing audio recording, WebSocket connections, and AI responses independently before integrating them

Real-time System Architecture: Developing a platform that processes data in real-time (voice input, AI responses, live chat)

What's next for Intervoice

We’re still exploring ways to make the voice integration fully functional by optimize streaming, improve reliability, and achieve smoother, more natural voice responses.

Expanding beyond interview prep to include resume optimization, LinkedIn profile enhancement, networking strategies, and personalized career roadmap suggestions

Built With

• claudeapi
• dart
• flutter
• tavily
• websocket