Nivan.AI Cloud Operations Platform

01 / OVERVIEW

Context & problem

Nivan.AI is a multi-cloud operations and governance platform for AWS and GCP. As the primary AI engineer, Adi develops agentic workflows and contributes across data, backend, and infrastructure layers. The system analyzes cloud telemetry and operational knowledge to support FinOps, security, reliability, and governance workflows.

Operational answers need to be grounded in current telemetry and retrieved knowledge, while respecting tenant boundaries and human authorization. Agent workflows must make tool access structured and keep proposed operations reviewable.

02 / SYSTEM ARCHITECTURE

How the pieces connect.

Conceptual architecture only. It intentionally omits proprietary implementation details, customer information, and operational data.

03 / INDIVIDUAL CONTRIBUTION

What I worked on.

• Architected LangGraph agents using OpenRouter-hosted models for multi-step reasoning, contextual analysis, and tool execution.
• Built root-cause workflows that correlate ingested AWS logs, metrics, and events into evidence-based findings.
• Implemented MCP servers that expose platform capabilities and operational data through structured tools.
• Developed RAG workflows and telemetry normalization pipelines.
• Contributed to human-approval workflows, PostgreSQL RLS multi-tenant isolation, and Kubernetes services.

04 / IMPLEMENTATION

Engineering details.

Operational answers need to be grounded in current telemetry and retrieved knowledge, while respecting tenant boundaries and human authorization. Agent workflows must make tool access structured and keep proposed operations reviewable.

• Ingested cloud telemetry is normalized into a shared data foundation.
• Retrieval workflows supply operational context to agent runs.
• LangGraph orchestrates analysis and MCP tool calls.
• Human-in-the-loop controls gate operational actions.
• The platform runs multi-tenant services across AWS and GCP and is designed around SOC 2 control requirements; this does not imply certification.

05 / RESULTS & VALIDATION

What the work supports.

• Built production-oriented agent, retrieval, and root-cause-analysis capabilities for a multi-cloud platform.
• Contributed from AI orchestration through secure data architecture and deployment.

Validation details are limited to evidence stated in the resume; no unreported datasets, accuracy claims, or benchmark comparisons are added.