Quizzy

Quizzy is an AI-powered virtual interviewer built to simulate personalized interviews through a robust integration of Retrieval-Augmented Generation (RAG), multi-modal analysis (vision, audio, text), and intelligent evaluation systems. It processes resumes and job descriptions, evaluates candidates in real-time, and generates detailed feedback, scoring, and job suggestions.Quizzy delivers an end-to-end mock interview experience that is hyper-personalized and interactive.

Built on Django, Quizzy orchestrates a modular architecture where core machine learning components are managed in a dedicated ML repository. These components are dynamically pulled and tracked using MLflow, DagsHub, and MLOps pipelines, ensuring reproducibility, scalability, and version control for both data and models.

---

Demo

Video Overview
Watch the full walkthrough of how Quizzy functions and performs an AI-driven interview:
YouTube Demo Link

---

Architecture and Features

Workflow

1. Input Collection: Users provide their resume, job description, and preferred interview time.
2. Resume Analysis:
   • Name and domain extraction
   • Resume-job description similarity using MXBAI embeddings (cosine similarity)
3. Interview Phase:
   • Dynamic question generation via Groq's LLaMA 70B model
   • Document retrieval from ChromaDB with Gemini embeddings
   • Real-time TTS (Edge TTS) and STT (Whisper)
   • Vision-based posture and emotion detection using MediaPipe and MobileNet
   • Profile summarization using HuggingFace model
4. Post Interview:
   • Score computation and evaluation report
   • Suggestions for improvement
   • Curated job recommendations from LinkedIn scraping based on candidate profile

ML and MLOps Stack

• Model Management: MLflow
• Experiment Tracking & Versioning: DagsHub
• Vision Models: MobileNet (transfer learned), MediaPipe
• Text Models: Groq (LLaMA 70B), Gemini Embeddings
• TTS/STT: Edge TTS and Whisper
• Resume Matching: MXBAI embeddings
• Document Store: ChromaDB

ML code and training pipelines are managed in a separate repository:
https://github.com/RijoSLal/Quizzy_MLOPS
Models and data are versioned and stored at:
https://dagshub.com/slalrijo2005/Quizzy_MLOPS

---

⚠️ Minimum System Requirements

• CPU: 12 threads (e.g., 4P + 4E cores, 2.0 GHz+)
• RAM: 6 GB
• GPU: Integrated, 48 EUs+, OpenGL 4.5+

⚠️ Systems below these specs may face performance issues or instability.

---

Deployment

To configure a production-ready self-hosted environment using NGINX and Cloudflare Tunnel, refer to my detailed blog post:
Self-hosting Django with NGINX & Cloudflare Tunnel

---

Local Development

Requirements

• Python 3.8+
• pip
• virtualenv

Setup

git clone https://github.com/RijoSLal/quizzy.git
cd quizzy
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows
pip install -r requirements.txt

Set Environment Variables

Create a .env file:

APIKEY=gemini_api_key
GROQ=groq_api_key
GROQQ=groq_api_key
GROQ_API_KEY=groq_api_key
DJANGO=django_secret_key

Run Tests

python manage.py test

---

Run the Server

python manage.py runserver

---

Docker Deployment

Build Docker Image

docker build -t quizzy .

Run with Environment File

docker run --env-file .env quizzy

Run Tests

docker run --env-file .env quizzy python manage.py test

---

Directory Structure

quizzy/
├── chroma_langchain_db/
├── interview/
|   ├── urls.py
|   ├── no_stream_camera_capture.py
│   ├── camera_capture.py
│   ├── resume_management.py
│   ├── retriever.py
│   ├── scrape.py
│   ├── speech.py
│   ├── vectordb.py
│   └── views.py
├── manage.py
├── quizzy/
│   ├── settings.py
│   └── urls.py
├── static/
|   ├── images/
│   └── styling/
├── templates/
│   ├── home.html
│   ├── interview.html
│   ├── resume.html
│   └── score.html
├── Dockerfile
├── .dockerignore
├── .env
└── requirements.txt

---

Contribution

Contributions are welcome. Please open an issue or submit a pull request if you would like to suggest improvements or contribute features. Ensure your contributions follow the project’s code standards and include relevant documentation.

---

License

This project is licensed under the MIT License.
See the LICENSE file for details.

---