AI Studio - AI Interview Simulator

AI Studio is an AI interview simulator consisting of a React Native (Expo) mobile frontend and a FastAPI Python backend. The platform provides real-time vocal analysis, role-specific question generation, and automated feedback using Large Language Models (LLMs) and local speech-to-text models.

🚀 Features

• Dynamic Question Generation: Leverages an LLM to generate interview questions based on standard behavioral banks or specialized technical roles.
• Audio Analysis: Analyzes audio responses for metrics like silence ratio, pitch variability (YIN algorithm), jitter, and speaking pace to calculate confidence and nervousness scores.
• Local AI Processing: Uses OpenAI's Whisper model for local speech-to-text transcription and Ollama (Qwen 2.5) for candidate evaluation.
• Modern Mobile UI: A React Native application featuring dark/light modes, audio metering visualizations, and structured feedback reports.

---

🛠️ Prerequisites

Before you begin, ensure you have the following installed on your machine:

• Python 3.8+
• Node.js 18+ & npm/yarn
• FFmpeg: Required by pydub and librosa for audio manipulation.
• Ollama: Required to run the local LLM evaluator.
  • Download from ollama.com
  • Pull the required model: ollama pull qwen2.5:7b-instruct-q4_K_M

---

💻 Backend Setup (FastAPI)

The backend handles audio processing, transcription, and LLM evaluation.

1. Install Required Packages

Navigate to your backend directory and install the following pip packages:

pip install fastapi uvicorn pydantic python-multipart openai-whisper torch librosa soundfile noisereduce scipy pydub requests

(Note: If you have an NVIDIA GPU, ensure you install the CUDA-enabled version of PyTorch for faster Whisper transcription).

2. Start the Local LLM Server

Ensure Ollama is running in the background with the correct model:

ollama serve
# Verify the model is available
ollama run qwen2.5:7b-instruct-q4_K_M

3. Run the Backend Server

Start the FastAPI server using Uvicorn. The --host 0.0.0.0 flag ensures the server is accessible from your mobile device/emulator on the same local network.

python -m uvicorn main:app --host 0.0.0.0 --port 8000 --reload

The backend API will be available at http://localhost:8000.

---

📱 Frontend Setup (React Native / Expo)

The frontend is an Expo-managed React Native application.

1. Install Required Packages

Navigate to your frontend directory and install the necessary npm dependencies:

# Core Expo and React Native
npm install expo react-native

# Navigation
npm install @react-navigation/native @react-navigation/native-stack react-native-screens react-native-safe-area-context

# Expo Hardware/UI libraries
npm install expo-av expo-camera expo-status-bar @expo/vector-icons

# Utils
npm install react-native-uuid

2. Configure Environment Variables

Create a .env file in the root of your frontend directory and add your backend's local IP address. Do not use localhost if you are testing on a physical device; use your computer's local IPv4 address (e.g., 192.168.1.X).

EXPO_PUBLIC_SERVER_URL=[http://192.168.1.](http://192.168.1.)X:8000

3. Run the Application

Start the Expo development server:

npx expo start

• Press a to run on an Android emulator.
• Press i to run on an iOS simulator.
• Scan the QR code with the Expo Go app on your physical device.

---

📂 Project Structure Overview

• Backend (/services):
  • main.py: FastAPI entry point and routes (/questions, /upload).
  • audio_processor.py: Cleans audio files, extracts acoustic metrics (jitter, pitch, confidence).
  • whisper_service.py & transcription.py: Manages the Whisper speech-to-text pipeline.
  • llm_evaluator.py: Constructs LLM prompts and parses JSON evaluation data.
  • questions_service.py: Context-aware prompt generation for dynamic questioning.
• Frontend (/screens & /utils):
  • WelcomeScreen.js, NoteScreen.js, PreferencesScreen.js: User onboarding and session configuration.
  • InterviewScreen.js: Core recording UI with live audio metering and animations.
  • ResponseScreen.js & uploadAudio.js: Batch processing and multipart form-data uploading.
  • FeedbackScreen.js: Renders the AI's final assessment and metric breakdowns.
  • ThemeContext.js & theme.js: Global dark/light theme state and styling constants.