Skip to content

Interviewer

Overview

Interviewer is an AI-powered interview practice application designed to help candidates prepare for real job interviews through voice-enabled mock sessions. Built with a local-first privacy approach, it runs entirely on your own hardware with no data ever leaving your machine.

The application generates dynamic, context-aware interviewer personas based on actual job descriptions and your resume, conducts multi-phase technical and behavioral interviews with real-time voice interaction, and provides comprehensive, actionable feedback.

Interviewer Application

Features

  • Voice-Enabled Interviews - Realistic interview simulation with speech-to-text (ASR) and text-to-speech (TTS)
  • Local AI Models - Runs entirely on your hardware via Lemonade Server with no cloud dependencies
  • Smart Document Extraction - AI-powered parsing of resumes and job descriptions
  • Dynamic Persona Generation - Creates tailored interviewer personas based on job/role
  • Comprehensive Feedback - Detailed performance analysis with actionable insights
  • Privacy-First - All data stored locally in JSON format
  • Cross-Platform Desktop App - Electron-based for Windows, macOS, and Linux

Setup

Prerequisites

  1. Install Lemonade Server by following the Lemonade Server Instructions
  2. Node.js 20+ - Download from nodejs.org
  3. Git - For cloning the repository

1. Clone and Install Dependencies

git clone https://github.com/lemonade-sdk/interviewer.git
cd interviewer

# Install Node.js dependencies
npm install

2. Start Lemonade Server

Start Lemonade Server using the desktop app or CLI:

# Using the desktop app - click the lemon icon (LemonadeServer.exe on Windows, lemonade-tray on macOS/Linux)
# Or using the CLI to start the HTTP server directly:
lemond

3. Pull Required Models

Interviewer uses multiple AI models for different capabilities. Pull the recommended models:

# LLM for interview logic and feedback (recommended: Qwen3-Coder-30B-A3B-Instruct-GGUF)
lemonade pull Qwen3-Coder-30B-A3B-Instruct-GGUF

# Whisper-Base for speech-to-text (automatic transcription)
lemonade pull Whisper-Base

# Kokoro for text-to-speech (natural voice synthesis)
lemonade pull kokoro-v1

4. Run Interviewer

# Development mode
npm run dev

# Or build for production
npm run build
npm run build:electron

Model Recommendations

Based on testing, the following models work well with Interviewer:

Capability Recommended Model
LLM (Interview Logic) Qwen3-Coder-30B-A3B-Instruct-GGUF
ASR (Transcription) Whisper-Base
TTS (Voice) kokoro-v1

Hardware Requirements: - Tested on AMD Strix Halo with 128GB RAM - Minimum 32GB RAM recommended for larger models - GPU acceleration recommended for faster responses

How It Works

Interviewer leverages Lemonade Server for all AI inference:

  1. Resume/Job Parsing - Uses LLM to extract key skills, requirements, and context
  2. Persona Generation - Creates an interviewer persona tailored to the job description
  3. Question Generation - Dynamically generates relevant technical and behavioral questions
  4. Voice Processing - Uses Whisper for transcription and Kokoro for natural voice synthesis
  5. Feedback Analysis - Provides detailed feedback on your responses with improvement suggestions

Architecture

Interviewer App (Electron + React)
        |
        | HTTP API (localhost:13305)
        v
Lemonade Server
        |
        +-- LLM (llama.cpp) --> Interview logic, feedback
        +-- Whisper ---------> Speech-to-text
        +-- Kokoro ----------> Text-to-speech

Troubleshooting

Model Loading Issues

If models fail to load, check that Lemonade Server is running: 

lemonade status

Audio Not Working

Ensure your microphone and speakers are properly configured in your system settings.

Slow Response Times

  • Ensure you have enough RAM (32GB+ recommended for Qwen3-Coder-30B)
  • Enable GPU acceleration if available
  • Close other resource-intensive applications

Repository

License

MIT License - see the LICENSE file for details.