The increasing reliance on cloud-native architectures, serverless computing, and artificial intelligence-driven systems has introduced new complexities in ensuring system dependability, resilience, and safety. Traditional reliability engineering approaches, while foundational, are often insufficient in addressing the dynamic, distributed, and failure-prone nature of modern cloud ecosystems. This research presents a comprehensive, theoretically grounded framework that integrates chaos engineering, machine learning-based reliability modeling, and human-centered safety principles to enhance system robustness across cloud-native and safety-critical domains, including healthcare and autonomous systems.

The study synthesizes interdisciplinary perspectives from cloud computing, dependability engineering, fault injection methodologies, and AI-based safety analysis. It explores how experimental fault injection, particularly through chaos engineering practices, can be combined with predictive analytics to proactively identify and mitigate system vulnerabilities. Furthermore, the research emphasizes the importance of realism in error injection, the role of serverless architectures in resilience testing, and the integration of human factors in safety-critical environments.

A qualitative, theory-driven methodology is employed to construct a unified framework that bridges gaps between cloud system resilience and safety engineering in domains such as healthcare. The findings suggest that integrating chaos engineering with machine learning enhances predictive fault detection, improves failure propagation understanding, and supports adaptive system recovery mechanisms. Additionally, the study highlights that human-centered design and error taxonomy integration significantly contribute to reducing systemic risks in critical infrastructures.

The proposed framework offers a novel contribution by aligning chaos engineering practices with AI-driven reliability assessment and safety assurance principles. It provides a scalable and adaptable approach for organizations seeking to build resilient, trustworthy, and high-performance systems in increasingly complex technological landscapes.