Engineered a comprehensive full-stack healthcare platform, SmartDent AI, for automated dental pathology detection and optimized patient recovery. Developed and optimized two deep learning models: a Disease Detection Model achieving 80% mAP@50 for 7 dental anomalies, and an FDI Tooth Numbering Model achieving 97% mAP@50. Implemented a custom FastAPI dual-model inference pipeline for simultaneous detection, segmentation, and automated mapping of pathologies to FDI tooth numbers using IoU logic

Engineered an end-to-end ANPR pipeline using YOLOv11 for vehicle detection, ByteTrack for multi-object tracking, and EasyOCR for license plate text extraction, optimized for real-time performance and occlusion handling with OpenCV

Created an automated video analysis pipeline for football matches using fine-tuned YOLO and ByteTrack to detect and track players, referees, and the ball. Implemented K-Means clustering for team assignment and proximity-based ball possession tracking.

Engineered a real-time traffic monitoring system using YOLO and ByteTrack for multi-class vehicle detection, tracking, and trajectory analysis. Developed a speed estimation algorithm and automated speeding violation detection across highway lanes.

An interactive 2D simulation of the Extended Kalman Filter Simultaneous Localization and Mapping (EKF SLAM) algorithm. Built using Python and Pygame, this project simulates a differential drive robot navigating an unknown environment, detecting landmarks, and building a map in real-time while estimating its own trajectory

Traditional object detection can identify the presence of a drill, but it cannot determine whether the drill is actually working. This project combines object detection, multi-object tracking, and motion analysis to classify each drill as ACTIVE or IDLE, then calculates the total drilling time.