ARVAS (Anomaly Recognition Video Analytics System) is an AI-assisted video anomaly detection system designed for surveillance applications. The system uses unsupervised machine learning to continuously analyze video streams and identify abnormal behavior in real time without requiring pre-configured rules or manual input. The system operates autonomously after installation, learning to distinguish between normal and abnormal behaviors across unlimited variations. It employs event clustering to group similar events together and enable user prioritization. ARVAS utilizes GPU acceleration to serve up to 200 cameras per server, making it scalable for deployments involving thousands of cameras. The platform is compatible with legacy camera sources, accepting relatively low frame rates and resolutions, and is not dependent on specific camera scenes. It can adapt to different environments across various sectors including security monitoring, facilities management, and critical infrastructure protection. ARVAS integrates with multiple video management systems including Milestone, Genetec, NxWitness, and Huawei. The system aims to reduce reliance on routine physical patrols and enable more efficient monitoring operations by highlighting anomalous events to operators in real time.