Autonomous AI Predictive Analytics Platform

End-to-end customer churn prediction project using the Telco dataset. Includes EDA, data preprocessing, Logistic Regression / Random Forest / XGBoost model comparison, SHAP explainability, and a production-ready prediction pipeline.

This project employs Logistic Regression for binary classification, to predict whether a borrower is capable of repaying a loan based on various financial and demographic factors.

ML-powered crop yield prediction for Maharashtra — Gradient Boosting (R²=0.977), SHAP explainability, MLflow experiment tracking, LLM-powered AI explanations, and interactive Streamlit dashboard.

LangGraph-based agent pipeline for code evaluation using static analysis, FAISS-powered RAG, and LLMs with SHAP-style explainability.

MultiModal Disaster Response System — BERT + ResNet-50 fusion for real-time disaster intelligence

Real-time PPG-based atrial fibrillation detection using Random Forest with feature-level explainability and interactive visualization.

Spatio-temporal graph deep learning for IPL T20 match outcome prediction. GAT player-interaction graph + BiLSTM + cross-attention Transformer. Ball-by-ball win probability, run forecasting & Player Impact Score with SHAP explainability.

The AI Loan Analyst is a sophisticated Streamlit-based web application designed to automate and enhance the loan analysis process for financial institutions. It combines data science, machine learning, and financial modeling to provide a complete loan portfolio management solution.

I built this application to allow users to input various clinical parameters and receive an instant prediction of whether a patient is likely to have CKD. The app is hosted on Streamlit community cloud for public access.

Built a machine learning model to predict telecom customer churn using classification techniques and SHAP explainability. Optimized performance through tuning and translated results into actionable customer retention insights.

FraudDetectAI is an advanced credit card fraud detection system built with XGBoost and Hybrid SMOTE Sampling (Oversampling + Undersampling). This project tackles highly imbalanced datasets, ensuring strong fraud detection accuracy while minimizing overfitting risks.