InterviewX

Offline AI-Powered Mock Interview Agent

InterviewX is a completely offline mock interview platform that simulates realistic interview conversations using state-of-the-art AI models. Practice your interview skills privately without any data leaving your computer.

✨ Key Features

• 🔒 100% Offline - All processing happens locally, no internet required after setup
• 🤖 AI-Powered Conversation - Natural interview dialogue using Qwen3-4B (4 billion parameter model)
• 🎤 Real-Time Speech Recognition - Advanced STT with Whisper and voice activity detection
• 🔊 Natural Voice Synthesis - High-quality text-to-speech using Kokoro TTS
• 📊 Performance Feedback - AI-generated evaluation and rating after each session
• 💼 Professional Interface - Clean, modern web UI with video conferencing experience
• 🖥️ Cross-Platform - Runs on Windows, macOS, and Linux

📁 Project Structure

InterviewX-Win-Max/
├── Frontend/          # React + TypeScript web interface
│   ├── src/          # React components, hooks, and types
│   └── README.md     # Frontend setup and development guide
├── Backend/          # Python AI engine and API server
│   ├── llm_engine.py    # Qwen3-4B conversation model
│   ├── stt_engine.py    # Whisper speech-to-text
│   ├── tts_engine.py    # Kokoro text-to-speech
│   ├── backend_engine.py # FastAPI server
│   └── README.md     # Backend setup and model guide
└── README.md         # This file

🛠 Technology Stack

Frontend

• Framework: React 19 with TypeScript
• Build Tool: Vite
• Styling: TailwindCSS
• Communication: WebSocket for real-time audio streaming

Backend

• Language: Python 3.9+
• Web Framework: FastAPI with WebSocket support
• AI Models:
  • LLM: Qwen3-4B (GGUF Q4_K_M, ~2.5 GB)
  • STT: OpenAI Whisper small (~466 MB)
  • TTS: Kokoro ONNX v1.0 (~60-120 MB)

🚀 Quick Start

Prerequisites

• Node.js 18+ and npm
• Python 3.9, 3.10, or 3.11
• RAM: 16 GB recommended for smooth LLM inference
• Storage: ~7 GB free space for AI models
• Hardware: Microphone and speakers/headphones

Setup Steps

1. Download AI Models

   The backend requires several AI model files (~3-3.2 GB total). See detailed instructions in the Backend README.

2. Setup Backend

   cd Backend
   python -m venv venv
   
   # Windows
   venv\Scripts\activate
   
   # macOS/Linux
   source venv/bin/activate
   
   pip install fastapi uvicorn websockets llama-cpp-python pywhispercpp mediapipe webrtcvad pyaudio sounddevice soundfile numpy

   For detailed backend setup including model configuration, see Backend README.

3. Setup Frontend

   cd Frontend
   npm install
   npm run build

   Then copy the build to Backend static folder:

   # Windows (from Backend directory)
   xcopy /E /I /Y ..\Frontend\dist static
   
   # macOS/Linux (from Backend directory)
   cp -r ../Frontend/dist/* static/

   For detailed frontend setup, see Frontend README.

4. Run the Application

   cd Backend
   python main.py

   The backend server starts on http://localhost:8000 and automatically opens your browser to the interview interface.

📖 Documentation

• Frontend Documentation - React UI setup, development, and architecture
• Backend Documentation - AI engines, model setup, and API reference

⚙️ System Requirements

• Operating System: Windows 10+, macOS 10.15+, or Linux (Ubuntu 20.04+)
• Memory: Minimum 8 GB RAM, 16 GB recommended
• Storage: ~7 GB for all AI models
• GPU: Optional (CPU-only operation fully supported)

🔧 How It Works

┌─────────────┐
│ User Speaks │
└──────┬──────┘
       │ Audio
       ▼
┌─────────────────┐
│ STT Engine      │ Whisper + VAD
│ (Speech-to-Text)│
└──────┬──────────┘
       │ Text
       ▼
┌─────────────────┐
│ LLM Engine      │ Qwen3-4B generates
│ (Interview AI)  │ interview questions
└──────┬──────────┘
       │ Response Text
       ▼
┌─────────────────┐
│ TTS Engine      │ Kokoro converts
│ (Text-to-Speech)│ text to speech
└──────┬──────────┘
       │ Audio
       ▼
┌─────────────┐
│ User Hears  │
└─────────────┘

End of Interview: The LLM analyzes the entire conversation and provides spoken feedback plus a performance rating (1-3 scale).

🎯 Usage

1. Start Interview - Click "Start Meeting" button
2. Grant Permissions - Allow microphone and camera access
3. Begin Interview - Say "start" or introduce yourself
4. Answer Questions - Respond naturally to AI interviewer's questions
5. End Interview - Click "End Call" when finished
6. Get Feedback - Receive AI-generated performance rating and feedback

📝 License

This project uses various AI models and frameworks, each with their own licenses:

• Qwen3: Alibaba Cloud (Apache 2.0)
• Whisper: OpenAI (MIT)
• Kokoro TTS: Community contributors
• FastAPI: Sebastián Ramírez (MIT)

---

Made with ❤️ for offline AI-powered interview practice