MY WORK
A collection of my Machine Learning, Deep Learning, Data Science, Analytics and Time-Series projects.
Back to PortfolioDeveloped a custom U-Net for pixel-level binary segmentation of dermoscopic skin lesions on 900 ISIC 2016 images, implementing encoder-decoder architecture, skip connections, image preprocessing, and BCE + Dice loss optimization.
An end-to-end Retrieval-Augmented Generation (RAG) system that answers questions from user-provided documents using semantic retrieval, FAISS vector search, and grounded text generation.
Deep learning-based image classification system that detects and classifies common solar panel defects from uploaded images.
Product analytics project exploring customer behavior, purchasing patterns, segmentation, revenue drivers and business insights using Python, SQL and Power BI.
Time-series forecasting project using LSTM to predict food delivery demand and capture historical demand patterns.
Deep learning-based fashion recommendation system using VGG16 feature extraction and cosine similarity for visually similar products.
NLP-based spam email classification system using TF-IDF feature extraction and a tuned Bernoulli Naive Bayes model, achieving high classification performance.
Developed a SARIMAX-based forecasting system to model historical demand, trends and seasonal patterns for inventory planning, with an interactive Streamlit dashboard.
AI-based histopathology image classification project using deep learning, transfer learning and Grad-CAM explainability for digital pathology applications.
Machine learning web application using XGBoost and Flask to detect potentially fraudulent credit card transactions.
Time-series forecasting project using Facebook Prophet to predict future Tesla stock prices from historical market data.
End-to-end machine learning pipeline for house price prediction using the Ames Housing dataset, including data preprocessing and feature engineering.