An Intelligent Cloud-Native GenAI Architecture for Project Risk Prediction and Secure Healthcare Fraud Analytics

Abstract

The rapid expansion of digital ecosystems, decentralized work environments, and high-velocity project delivery pipelines has amplified the need for intelligent, real-time risk forecasting and cyber-fraud detection systems. Traditional predictive models, while effective in static contexts, struggle to adapt to dynamic, cloud-first operational landscapes that produce heterogeneous, high-frequency data. This paper proposes a cloud-native generative AI architecture integrating Gaussian Process Regression (GPR) and Multilayer Perceptron (MLP) hybrid models for real-time project risk forecasting and cyber-fraud detection. The architecture leverages containerized microservices, event-driven pipelines, and distributed vector databases to support continuous learning, scalable inference, and generative scenario synthesis. Generative AI components—implemented through transformer-based models—simulate evolving risk patterns, create synthetic training datasets, and enhance anomaly detection capabilities in cybersecurity contexts. The hybrid GPR–MLP approach improves predictive stability by combining probabilistic uncertainty quantification from GPR with the nonlinear learning capacity of deep MLP networks. Results from simulated enterprise cloud environments indicate significant improvements in detection accuracy, risk forecast reliability, and system resiliency under high-load conditions. The proposed model demonstrates superior performance in identifying subtle behavioral anomalies associated with financial fraud, access privilege abuse, and insider threat activities. For project risk forecasting, the architecture captures cross-dependent operational signals such as cost deviations, schedule drifts, and resource bottlenecks with improved lead time and interpretability.