← Back to Capstones

cti-260 --project

AI-Driven Proxmox Automation in a Converged Data Center Lab

This capstone demonstrates a governed automation pattern for a virtualized lab environment using Proxmox, scoped service access, MCP-style tooling, logging, and repeatable infrastructure operations. The project shows how I connect virtualization, automation, security boundaries, and operational troubleshooting into a controlled infrastructure workflow.

Overview

A governed AI-operations pattern connecting Claude Desktop and VS Code to a Proxmox MCP server for scoped, logged, secret-free infrastructure actions across Proxmox VE, Cisco FlexPod/UCS, and NetApp.

Problem

Hiring managers are wary of letting AI tools touch live infrastructure — unscoped access, missing audit trails, and secret exposure are the #1 concern. This project designs a pattern where AI can assist with infrastructure operations without those risks.

Goals

  • Let an AI assistant (Claude Desktop / VS Code) query and act on a Proxmox lab safely
  • Guarantee every AI action is scoped, logged, and secret-free
  • Keep a human validation step before and after each action
  • Extend the pattern across Proxmox VE, Cisco FlexPod/UCS, and NetApp

Design Approach

workflow> Observe → Query → Analyze (AI) → Act → Validate
# every AI action is scoped, logged, and secret-free

A Proxmox MCP server (Ubuntu LXC) exposes a scoped API surface with service-account separation, so the AI layer never holds broad credentials. Every action is logged and traceable, and a human reviews before and after execution.

  • Service-account separation over shared credentials, so AI access can be scoped and revoked independently
  • Full logging and traceability chosen over speed/convenience, to keep every AI action auditable
  • Human validation gates before and after action execution, rather than fully autonomous operation

Technical Components

  • Claude Desktop
  • VS Code
  • Proxmox MCP server (Ubuntu LXC)
  • Proxmox VE
  • Cisco FlexPod / UCS
  • NetApp

Business / Operational Value

Directly addresses the #1 hiring-manager concern about "AI touching infrastructure" — every AI action is scoped, logged, and secret-free, with a human validation step before and after. Full system design is documented as a downloadable PDF; this is a lab/capstone design, not a production deployment for a client.

Skills Demonstrated

  • Governed AI/automation tool integration (MCP-style scoped tooling)
  • Service-account separation and least-privilege API access
  • Logging, traceability, and audit-ready operations
  • Virtualization administration (Proxmox VE, Cisco FlexPod/UCS, NetApp)
  • Human-in-the-loop validation design for automated infrastructure actions

Want to walk through the design decisions?