Faiyaz ali shaik

Cloud Finops engineer

Abstract
As enterprises spread workloads across AWS, Azure, and Google Cloud, cost management becomes fragmented. Each provider uses its own billing schema, tagging conventions, discount constructs, and reporting cadence, so cost visibility is delayed and cost accountability is unclear. Traditional FinOps tooling relies on static dashboards and manual review, which detect overspend only after it has already occurred.

This paper presents an AI-driven FinOps pipeline that unifies cost governance across the three providers. The pipeline ingests native billing data (AWS Cost and Usage Reports, Azure Cost Management exports, and GCP BigQuery billing export) and normalizes it into a common schema aligned with the FinOps Foundation's FOCUS specification. Machine-learning models then run on the normalized data. Time-series anomaly detection flags unexpected spend, and a rolling-window repeat-offender analysis separates recurring cost drivers from one-off spikes. Forecasting models project spend per account, project, and service. A large language model layer turns anomalies into plain-language root-cause explanations by correlating cost signals with resource metadata, tagging gaps, and deployment changes, so engineers get context they can act on instead of raw alerts.

The pipeline goes beyond reporting by adding a policy-guarded recommendation and remediation loop. It suggests rightsizing, idle-resource cleanup, and commitment optimization (Savings Plans, Reserved Instances, and committed use discounts). Any automated action requires human approval, so governance is preserved while manual effort drops. Insights are delivered through role-based dashboards and alerts for engineering, finance, and leadership.

The proposed architecture shows how AI can shift cloud financial management from reactive reporting to proactive, explainable, and semi-autonomous optimization in multi-cloud environments. The work also discusses design challenges such as cross-provider data normalization, shared-cost allocation, model explainability, and the trust boundaries for AI-initiated actions.

Keywords: FinOps, Multi-Cloud, AWS, Azure, GCP, Cost Anomaly Detection, Generative AI, FOCUS Specification, Cloud Cost Optimization