MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641048618 A) filed by Dr. Dhiyanesh; Athithya S. A; Vivin K S; Eedpuganti Yagna Sai Harshith; and J. Princeton Vishal on April 16, 2026, for Self Evolving Autonomous Intelligence System With Digital Twin Federated Learning & Cognitive Ai Int.
Inventors include Dr. Dhiyanesh; Athithya S. A; Vivin K S; Eedpuganti Yagna Sai Harshith; and J. Princeton Vishal.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: The present invention discloses JARV1S-X, a self-evolving multi-model artificial intelligence system that integrates six interdependent Al models to achieve continuous, adaptive, and explainable intelligence. Existing Al systems are predominantly static — they are trained once and deployed without mechanism for autonomous adaptation, making them vulnerable to concept drift, data distribution shifts, and changing real- world conditions. JARVIS-X addresses this fundamental limitation by combining online machine learning via HalfSpaceTrees (River), Federated Averaging (FedAvg) across distributed edge nodes with dualdetector concept drift monitoring using ADWIN and Page-Hinkley algorithms, digital twin-based future state simulation, a GPT-powered Explainable Al (XAI) layer, a cognitive JARVIS interface for natural language and voice interaction, and a personalization engine for user-adaptive behavior. The system achieves real-time anomaly detection with sub-500ms latency, privacy-preserving distributed learning across organizational boundaries without raw data exchange, and 6-24 hour predictive horizon with quantified risk scores. The proposed architecture enables self-monitoring, self-correcting, and selfexplaining Al operation across healthcare, finance, manufacturing, education, cybersecurity, small cities, and autonomous vehicle domains. JARVIS-X represents a paradigm shift from static Al models to a living, continuously evolving intelligence platform capable of operating reliably in non-stationary, highdimensional, and privacy-sensitive real-world environments.
Disclaimer: Curated by HT Syndication.
