Sanket Dhonde

An AI-powered monitoring system for real-time safety and threat detection. Computer vision models analyze live video and images to identify suspicious activity, with automated alerts notifying users during potential risks — built as a full pipeline from detection through analysis to real-time response.

A deep learning-based system to detect manipulated (deepfake) images, training CNN architectures such as ResNet and DenseNet on benchmark datasets. Applied CBAM attention and Grad-CAM for interpretability, and explored One-Class Classification (Autoencoders, Deep SVDD) for detecting fakes with limited training data.

A comprehensive AI-driven healthcare platform that combines disease prediction, OCR-based prescription reading, medicine recommendations, digital health records, and intelligent healthcare assistance. The platform integrates multiple AI models to improve accessibility and patient care.

A research-driven AI system that detects phishing websites using URL heuristic analysis and classifies spam emails through Natural Language Processing. The work was published in an international journal and demonstrates practical cybersecurity applications of Machine Learning.

Developed a fraud detection model using supervised Machine Learning techniques on highly imbalanced transaction datasets. Applied feature engineering, preprocessing, and model evaluation to accurately identify fraudulent activities while minimizing false positives.