World Conference on Information Technology, Data Analytics, and Intelligent Systems (WCITDAIS-2026)

Digital health systems increasingly rely on artificial intelligence and machine learning to enhance clinical decision-making, diagnostic capabilities, and patient outcomes. However, the growing adoption of AI also introduces significant cybersecurity, privacy, and model-integrity challenges. This keynote will explore strategies for building secure, trustworthy, and resilient AI systems for digital health. It will discuss emerging threats such as adversarial machine learning, along with practical approaches including adversarial training, federated learning, privacy-preserving techniques, real-time anomaly detection, and robust model validation. The session will also address regulatory considerations, data heterogeneity, explainability, and the importance of human oversight. Drawing on research and practical perspectives, the keynote will highlight how cybersecurity and clinical effectiveness can be addressed together to support the responsible deployment of intelligent healthcare technologies.

My proposed presentation is titled Forecasting for Everyone: Building a Self-Serve Forecasting Platform. The session will explain how AI agents can help business users prepare data, select suitable forecasting approaches, understand results, and make informed decisions while maintaining human oversight, reliability, and transparency.

Artificial intelligence systems often perform well during experimentation but encounter significant challenges when deployed in production environments. Real-world AI applications must process changing data, integrate with complex enterprise platforms, operate securely, and continue delivering dependable results despite model drift, infrastructure failures, unexpected inputs, and evolving business requirements. This keynote examines the engineering principles required to transform AI prototypes into reliable, observable, and resilient production systems. It presents a practical framework covering data and model validation, end-to-end observability, automated failure detection, drift monitoring, explainability, security, governance, and human oversight. The session also explores how metadata-driven controls and intelligent agents can monitor pipelines, identify anomalies, assess downstream impact, recommend corrective actions, and support safe recovery. Examples from healthcare, financial services, cloud data platforms, and enterprise applications will demonstrate how organizations can design AI systems that remain trustworthy under real operational conditions. Attendees will learn how to establish service-level objectives, monitor data and model behavior, implement graceful degradation, maintain audit evidence, and build recovery mechanisms. The keynote provides a structured approach for moving beyond successful AI demonstrations and engineering intelligent systems that organizations can confidently operate at scale.

Attackers targeting federal critical infrastructure are using AI to find vulnerabilities, build adaptive exploits, and dodge detection faster than any human analyst can respond. Prompt injection, model poisoning, automated reconnaissance, and adversarial manipulation have opened attack surfaces that signature-based and rule-based defenses were never designed to handle. Energy grids, water systems, transportation networks, financial infrastructure, healthcare systems, and defense installations all depend on defenses that still wait for a known pattern to appear before they act.