E-Commerce Sales Analysis

October 2025

This project focuses on analyzing e-commerce sales data to uncover insights on revenue, customer behavior, and product performance. Using Python and data visualization libraries, it provides an interactive way to study trends and KPIs.

Key Features:

• 📊 Data cleaning, processing, and visualization using Pandas and Matplotlib
• 📈 Sales forecasting and trend analysis
• 🧠 Insight generation from real-world datasets

Tech Stack: Python · Pandas · Matplotlib · Data Visualization · Analysis

News Article Classification

January 2025 - June 2025

This project developed a multi-input AI system that classifies news articles into four categories: World, Sports, Business, and Sci/Tech. It supports inputs via plain text, PDFs, images (OCR), and URLs, using advanced NLP techniques.

Key Features:

• 📄 Multi-format input: text, PDF, image, URL
• 🧠 Built using Bidirectional LSTM with dropout layers for reduced overfitting
• 🔍 Used ConceptNet Numberbatch for semantic word embeddings
• 🎯 Achieved 86.55% validation accuracy through hyperparameter tuning

Tech Stack: Python · Pandas · Machine Learning · NLP · Word Embeddings

Potato Leaf Disease Detection

January 2025 - February 2025

This project leverages deep learning to classify potato leaf diseases into Healthy, Early Blight, or Late Blight using a Convolutional Neural Network (CNN) model. It features a web-based interface powered by Streamlit, allowing users to upload images and get instant predictions with confidence scores.

Key Features:

✅ Upload potato leaf images for disease detection

✅ Real-time classification with confidence scores

✅ Optimized TensorFlow Lite model for efficient inference

✅ Lightweight Streamlit UI for easy accessibility

Tech Stack:

Python (Backend & Model Integration)
TensorFlow Lite (Optimized Model Inference)
Streamlit (Web UI)
NumPy & PIL (Image Processing)

🔗 Try the Web App

MNIST Handwritten Digits Recognition using PyTorch

July 2024 - September 2024

The project focuses on building and training a machine learning model for handwritten digit classification using the MNIST dataset. The goal is to classify digits (0-9) based on 28x28 pixel grayscale images. Using Python and PyTorch, the project involves loading the dataset, preprocessing images, defining a neural network, and optimizing the model using Stochastic Gradient Descent (SGD). The project demonstrates the use of SoftMax for classification, cross-entropy for error measurement, and essential machine learning workflows like batching and backpropagation.

Personal Portfolio Website

July 2024 - August 2024

This portfolio website, developed using HTML and CSS, serves as a showcase of my work as a software engineer. It highlights my experience in banking and management, as well as key projects in project management, cybersecurity, and software development. The site demonstrates my technical skills in web development, offering a professional platform for potential employers and collaborators to explore my projects, skills, and career journey.