All work
AI / LLM Engineering

AI Network Audit Agent

An AI agent that adopts a "Principal Network Architect" persona to autonomously audit a Cisco Meraki-managed network via API, producing prioritized, actionable remediation plans with generated topology diagrams.

6 (security, wireless, switching, SD-WAN, monitoring, governance)

Audit domains covered

Finding → Risk → Remediation → Priority

Output format

CLI pipeline + conversational agent

Usage modes

The Problem

Comprehensive network audits (security posture, wireless design, switching hygiene, SD-WAN health, org-level configuration hygiene) are valuable but time-consuming to do thoroughly and consistently by hand, and easy to do shallowly under time pressure.

My Approach

  • Designed the agent around an explicit expert persona with real authority framing — evaluating every finding against established vendor best-practice/validated-design guidance — and a strict output format: Finding → Risk → Remediation → Priority, so results are consistently structured and immediately actionable instead of a wall of raw data.
  • Scoped full API-driven coverage across every major domain of the platform: security (firewall, IDS/IPS, content filtering, VPN, threat protection), wireless (SSID design, RF profiles, client isolation), switching (VLAN segmentation, spanning tree, port security, QoS), SD-WAN (uplink config, failover, traffic shaping), monitoring (alerting, syslog, SNMP), and organization-level hygiene (admin roles, API access, licensing).
  • Selected a concrete technical stack for turning raw API data into a useful deliverable: a diagramming library with native vendor icon support for auto-generated topology diagrams (plus Markdown-native diagrams for quick viewing), and a templated HTML→PDF pipeline for a polished, scored findings report.
  • Designed the tool to be usable two ways — as a standalone CLI/data-collection pipeline, or as a conversational AI-agent skill — so the same audit engine serves both an automated batch use case and an interactive "ask a follow-up question" use case.

Stack

AI/Agent Design

Persona-driven prompt engineeringStructured output designTool-use/agentic workflows

Data Collection

PythonOfficial Meraki Dashboard API SDK (async-capable)

Reporting/Visualization

diagrams (Python topology-diagram library)MermaidJinja2 + WeasyPrint (HTML → PDF)pandas/Excel

Practices

Network engineering best-practice frameworks (validated design methodology)Structured audit/reporting formats

Skills Demonstrated

  • ▸AI agent and persona design for a technical, high-stakes domain
  • ▸Deep network engineering domain knowledge (security, wireless, switching, SD-WAN, monitoring, org governance)
  • ▸Turning a large, unstructured API surface into a consistently formatted, decision-ready deliverable
  • ▸Designing tools that work equally well as automation and as an interactive assistant