Uttam1695/Spam-Shield-AI

SpamShield AI: Advanced LLM Spam Classification System 🛡️

Welcome to SpamShield AI, an intelligent, end-to-end spam classification web application. This project demonstrates the implementation of a custom-built Large Language Model (LLM) utilizing a transformer architecture to classify text messages (SMS/Emails) into Spam or Ham (Not Spam).

Every component of this pipeline—from data preprocessing and model architecture to the modern web interface—was engineered entirely from scratch.

🚀 Features

• —Custom Transformer Architecture: Implemented entirely in PyTorch, leveraging attention mechanisms optimized for natural language processing.
• —High-Performance Inference: Real-time classification of text streams via an integrated Streamlit frontend.
• —Modern User Interface: A responsive, beautifully designed frontend using Streamlit with custom CSS injection featuring glassmorphism and animated backgrounds.
• —Robust Training Pipeline: Includes custom PyTorch Datasets/DataLoaders, automated validation loops, and learning rate scheduling (detailed in the Jupyter Notebook).

🧠 Model Architecture & Training

The core of SpamShield AI is the review_classifier.pth model. It was trained on the benchmark SMS Spam Collection dataset.

Key Highlights:

1. 1.Data Engineering: Processed over 5,500 SMS messages, handling class imbalances and applying advanced tokenization strategies.
2. 2.Custom Embeddings: Built custom word and positional embeddings to capture deep semantic meaning and sequence context.
3. 3.Loss & Accuracy Tracking: The training process achieved exceptional accuracy, significantly reducing false positives (as visualized in accuracy-plot.pdf and loss-plot.pdf).
4. 4.State Dict: The trained weights (review_classifier.pth) are loaded seamlessly into the Streamlit application for low-latency predictions using @st.cache_resource.

💻 Tech Stack

• —Deep Learning: PyTorch
• —Backend & Frontend: Python, Streamlit
• —Data Processing: Pandas, NumPy
• —Environment: Jupyter Notebook (for model research and development)

🛠️ Installation & Setup

To run this application locally on your machine, follow these steps:

Prerequisites

Make sure you have Python 3.8+ installed.

1. Clone the repository

git clone https://github.com/yourusername/Spam_Classifier_using_LLM_from_scratch.git
cd Spam_Classifier_using_LLM_from_scratch

2. Install Dependencies

Install the required Python packages (it is recommended to use a virtual environment):

pip install -r requirements.txt

3. Run the Application

Start the Streamlit application:

streamlit run app.py

The interface will automatically open in your default web browser (typically at http://localhost:8501).

📸 Screenshots & UI Design

The web interface was meticulously designed to provide a premium user experience. It features:

• —Glassmorphism: Sleek, transparent UI elements that blur the dynamic background.
• —Real-time Feedback: Animated loading states and dynamic result cards that adapt based on the prediction confidence.
• —Responsive Layout: Flawless execution across desktop and mobile devices.

📈 Future Enhancements

• —API Rate Limiting: Implementing Redis to manage request throttling.
• —Model Quantization: Reducing the model size for edge deployment without sacrificing accuracy.
• —Multi-lingual Support: Expanding the vocabulary to detect spam in multiple languages.

Designed and engineered by Uttam Kumar Mahato