Dandelion User

I'm building AI systems that turn machine learning models into real-time applications, APIs, and dashboards. My work spans fraud prevention, healthcare analytics, computer vision, and multilingual emergency assistance.

At Cyberpace Foundation, I built an AI-based toll detection and fraud prevention system across 99 toll plazas using YOLOv11, OpenCV, EasyOCR, Random Forest, PostgreSQL, FastAPI, React, and TypeScript. The system provides real-time alerts and case investigation through REST and WebSocket APIs.

I built AuraCheck, a Flask application that predicts stress, anxiety, and depression severity from psychological assessments. After evaluating six machine learning models on a 5,393-feature dataset, I selected XGBoost, achieving 97.9% validation accuracy.

I also built Sahaayak AI, a multilingual emergency assistant supporting five languages and six emergency intent categories. It combines language detection, translation, TF-IDF classification, spaCy entity extraction, geolocation, and Google Maps to help users find nearby emergency services.