Toward Resilient AI Systems: Adaptive Cyber Defense and Privacy-Preserving Intelligence at Scale

Dependable AI requires security, privacy, scalability, and sustainability. This talk presents three research pillars: adaptive cybersecurity, federated learning for privacy-preserving edge intelligence, and hierarchical RL for autonomous systems. This talk presents a unified framework combining Hierarchical RL for coordination, RAG for knowledge retrieval, and MCP for tool execution. Drawing on IEEE research, this provides practical guidance for building dependable, cost-effective AI systems across various industries.