Projects
Knee Osteoarthritis Severity Prediction
Developed a web application that predicts the severity of knee osteoarthritis using Convolutional Neural Networks (CNNs). The model analyzes X-ray images to classify the severity level, aiding medical professionals in faster and more accurate diagnoses.
Yelp Data Translation to ArangoDB
Designed and implemented a data translation pipeline that transforms Yelp’s extensive JSON dataset into a graph-based ArangoDB structure. This conversion improves query efficiency, data relationships, and storage optimization, making it easier to extract insights from business reviews.
Anomaly Detection and Root Cause Analysis in Aviation Data
Led the development of a machine learning pipeline using XGBoost to detect anomalies in aviation flight data, achieving a 60% AUC score. Integrated Explainable AI techniques like SHAP and LIME to provide interpretability. Additionally, fine-tuned a Large Language Model (LLM) on aviation manuals and maintenance logs, enabling real-time root-cause analysis for detected anomalies.
Product Recommendation System
Built a recommendation system leveraging Amazon’s 373GB review dataset by transforming JSONL data into PostgreSQL RDS and embedding product data using OpenAI models. Implemented scalable vector similarity searches with Pinecone, optimizing product recommendations. Additionally, developed and deployed a chatbot using Flowise, integrating PostgreSQL and Pinecone for real-time, tailored product suggestions.
