🧠 My Interview Bot - Behavioral Interview Practice App

This is a local AI-powered interview practice tool that helps you improve your behavioral interview skills. It uses Ollama for local LLM inference and Whisper for speech-to-text transcription, allowing you to speak your answers and get instant feedback from an AI hiring manager.

---

🚀 Features

• 🎤 Voice-based answers using Whisper for transcription
• 🧑‍💼 Realistic behavioral interview questions tailored for any profession the user inputs
• 💬 AI feedback focused on clarity, structure, and communication style
• 💾 Local storage of interviews — all interview data (video files, transcripts, and feedback) are stored locally in the uploads/interviews/ directory.
  Each interview is organized in its own subfolder, and you can manage (view or delete) them directly from the MyInterviews page in the app.
• ⚙️ Customizable model selection
• 🔒 Runs fully locally (no cloud or API key required)

---

🧩 Prerequisites

Make sure you have the following installed before running the app:

Tool	Description	Install Command / Notes	Approx. Space
Java 17+	Required to run the Spring Boot app	Download Java	~300 MB
Ollama	Local LLM runtime	Install Ollama	~1.5 GB
AI model	Check out the available models at Ollama library	Example: ollama pull phi3	Depends on the model
Python 3 + pip	Required for Whisper	Install Python	~500 MB
FFmpeg	Required by Whisper for audio processing	macOS: brew install ffmpeg Ubuntu: sudo apt install ffmpeg	~200 MB
Whisper	Speech-to-text transcription	pip install -U openai-whisper	~1.5 GB

💡 Recommended Setup For the best experience, use a Mac with Apple Silicon (M1/M2/M3) or a machine with a dedicated GPU. These devices handle local LLM inference and Whisper audio processing much faster and more efficiently.

The app is also optimized for Google Chrome, which provides the most reliable camera, microphone, and recording support compared to other browsers.

---

⚙️ Installation & Setup

1. Clone this repository:

   git clone https://github.com/alanquintero/myInterviewBot
   cd myInterviewBot

2. Configure the AI model in application.properties (llama3.1:8b is the default model):

ai-model=llama3.1:8b

3. Run the app with Spring Boot:

    mvn spring-boot:run

4. Open the app:

5. Visit http://localhost:8080 in your browser.

---

🎙️ Troubleshooting

Issue	Possible Cause	Solution
🎧 No mic input / camera access	Browser permissions or wrong device selected	Check browser settings → Allow mic and camera, and ensure the correct devices are selected
❌ Cannot install Whisper	Make sure Python version is compatible with Whisper	Check Whisper official site openai-whisper
❌ “Model not found” error	You haven’t pulled the model	Example, run ollama pull phi3
🐍 pip command not found	pip not installed	macOS/Linux: sudo apt install python3-pip or brew install python3
🐢 Slow transcription	Whisper base model is large	Try smaller Whisper models (like tiny or base) (Not yet available in the app)

---

⚙️ Configuration

Developers can configure the AI model in application.properties:

ai-model=llama3.1:8b

You can replace llama3.1:8b with another Ollama model, such as mistral, phi3, or any other model available locally.

Or, when running the app, you can override any property like this:

mvn spring-boot:run -Dspring-boot.run.arguments="--ai.model=phi3"

Example of running a JAR overring properties:

java -jar myinterviewbot.jar \ --ai.provider=ollama \ --ai.model=phi3:latest

Example of running app in other port:

mvn spring-boot:run -Dspring-boot.run.arguments="--server.port=9090"

---

🧩 Choosing the Right Ollama Model

Picking the right model depends on what you care about most — speed, accuracy, or resource usage. Here’s a quick guide:

💻 Laptop Type	🧠 Recommended Model	📦 Approx. Size	✅ Benefits	⚠️ Limitations
8–12 GB RAM	Phi-3 Mini (3.8B)	~2 GB	🟢 Extremely fast 🟢 Small download 🟢 Great for quick Q&A and light tasks	🔴 Limited reasoning depth 🔴 Not ideal for long conversations
8–12 GB RAM	Mistral 7B	~4 GB	🟢 Smart and efficient 🟢 Handles follow-ups better than Phi-3 🟢 Good general-purpose model	🔴 Slightly robotic tone 🔴 Less consistent on complex logic
16–18 GB RAM	Llama 3.1 (8B)	~5 GB	🟢 Excellent reasoning 🟢 Natural, human-like answers 🟢 Great for behavioral interview simulation	🔴 Slightly slower startup 🔴 Requires quantized version for best speed
16–18 GB RAM	Gemma 2 (9B)	~6 GB	🟢 Balanced quality and speed 🟢 Friendly conversational tone 🟢 Efficient on Apple Silicon	🔴 Can occasionally repeat or overexplain
24+ GB RAM / M3 Pro–Max	Llama 3.1 (13B)	~8–9 GB	🟢 High-quality, detailed reasoning 🟢 Handles multi-turn interviews beautifully 🟢 Very consistent and coherent	🔴 Slower on smaller laptops 🔴 Heavy model
Server / Multi-GPU Setup	Llama 3.1 (70B)	~40–45 GB	🟢 Near GPT-4 quality 🟢 Exceptional reasoning and memory 🟢 Ideal for research or production AI agents	🔴 Requires 64 GB+ RAM or GPU cluster 🔴 Very slow download / load

Check available models at Ollama library.

Example of getting a model:

ollama pull llama3.1:8b

Show all the models that are currently installed on your machine:

ollama list

---

🚀 How It Works

1. Enter a Profession and click Generate Question to receive a tailored interview prompt.

   1.a Enter your own question or choose one from the most common interview questions.

2. Click Ready when you are ready to answer the question.

3. Record your answer directly in the app.

4. Review your video after recording.

5. Click Generate Feedback to analyze your response.

6. Instantly get a Transcript and AI-powered Feedback and Evaluation on your performance.

7. Visit the My Interviews page anytime to view all your past practice sessions.

8. Settings: You can customize the app from the Settings page, where you can:

   • Select the AI model to use
   • Choose between audio-only or video + audio recording
   • Set your default profession
   • Toggle visibility for question category and difficulty

9. System Checks: The app automatically checks your system hardware (camera, microphone, and processing capability).

10. If your device doesn’t meet the minimum requirements, you’ll see a system alert warning that performance might be limited.

---

🧭 Roadmap / Future Development

🚀 Core Features

• 🧭 Conversation history — track progress over time
• 💻 Custom Questions — enter the question you want to practice
• 🎯 Scoring system — rate clarity, confidence, relevance, etc.
• 📄 Resume Interview Mode — tailor questions based on uploaded resume
• 💻 Technical Interview Mode — technical questions
• 🤝 Mock Interview Mode – A session to practice common questions asked in job interviews
• 📊 Progress Analytics — graphs for improvement over time
• 🤖 Add more AI providers (e.g., OpenAI GPT-4, Claude, Gemini)
• 🗣️ Text-to-Speech for AI questions and feedback
• 🤖 Add support for smaller Whisper models (e.g., tiny, base)
• 🎨 Improved UI/UX — modern dashboard, light/dark mode, analytics
• ⚙️ Settings fully customizable from the app — change AI provider, change AI model, enable or disable text-to-speech, change recording time, only audio interview, etc.

🧪 Experimental Ideas

• 🧠 Feedback memory — personalized tips based on past sessions
• 💬 AI interviewer personalities (strict, friendly, technical)
• 🗂️ Integration with Google Drive or Notion for saving feedback

---

🧰 Tech Stack

• Backend: Java 17, Spring Boot
• Frontend: HTML, JavaScript, CSS, Bootstrap
• AI: Ollama, Whisper (speech-to-text), FFmpeg (audio)
• Build Tool: Maven

---

🧑‍💻 Contributing

Contributions are welcome! If you’d like to help add features, fix bugs, or improve documentation:

1. Fork the repo
2. Create a feature branch
3. Submit a pull request

---

📜 License

MIT License — free to use, modify, and distribute.