Governed Cloud AI for Financial Services: MLOps, Foundation Models, and Trustworthy AI Analytics across AWS, Azure, and Google Cloud
Main article
Abstract
This paper proposes an integrated reference architecture and a governance-first deployment perspective for cloud-based artificial intelligence in financial services. We argue that the strategic value of cloud AI for banks, insurers, and capital-market firms now depends less on the choice of provider and more on the institution's ability to operationalize three jointly mature capabilities: machine-learning operations (MLOps) as a disciplined production lifecycle, foundation-model access patterns that preserve confidentiality and explainability, and trustworthy AI analytics aligned with emerging governance regimes including the EU AI Act, the NIST AI Risk Management Framework, and ISO/IEC 42001. Drawing on a comparative analysis of Amazon Web Services, Microsoft Azure, and Google Cloud Platform, and a survey of governed deployments across 142 financial institutions in Europe, North America, and Asia, the paper develops a five-lane reference architecture that separates the data plane, the MLOps lane, the foundation-models lane, the governance lane, and the application plane. We map provider capability profiles across six dimensions, document F1 performance gains of 16-25 percentage points when rule-based controls are replaced with governed MLOps plus explainable AI on fraud detection, credit-risk modeling, anti-money laundering, and customer service tasks, and quantify the Pareto trade-off between inference latency and composite governance score across six deployment topologies. The paper concludes with a research and practitioner agenda emphasizing multi-cloud governance, confidential computing for sensitive analytics, foundation-model evaluation discipline, and the embedding of trustworthy-AI controls into the institution's existing model-risk-management apparatus.
