AI-Augmented Troubleshooting

Traditional troubleshooting scripts only work if you already know the exact command to run. I extend this by integrating intent classification and NLP models into my automation stack. That means I can type “check neighbors on Core01” and the framework knows exactly which CLI commands to run, how to parse the output, and how to summarize the results.

My Approach

• Natural Language → CLI Mapping – intent models classify plain English queries and translate them into safe, vendor-specific commands.

• Structured Parsing – command outputs are parsed into structured JSON for downstream correlation and reporting.

• Guided Next Steps – AI suggests what to check next if anomalies are detected, escalating intelligently through playbooks.

• Read-Only by Default – configuration changes are always gated by explicit human approval.

Agentic Tie-In

I build troubleshooting routines as agentic workflows:

• The system runs diagnostics automatically, interprets results, and decides the next logical command.

• Failed branches trigger alternate playbook paths without manual intervention.

• Engineers stay in the loop — AI suggests, humans approve.

Advancing Further

I continue to expand methodology toward:

• Cross-Vendor Abstraction – same intent (e.g., “show neighbors”) translates seamlessly across Cisco, Juniper, Arista, etc.

• Feedback Loops – user corrections feed back into the model to improve intent accuracy.

• Full Agent Integration – combining event correlation, diagnostics, and packet capture into a single AI-driven workflow.

Why It Matters

AI isn’t here to replace engineers — it’s here to amplify them. By augmenting my troubleshooting stack with intent-driven workflows, I enable faster diagnostics, safer automation, and engineer-grade insights on demand.