I'm a network and security engineer. I've spent close to twenty years keeping a large network running and, more and more, defending it.

My days are split two ways. One half is firewalls, wireless and identity systems. The other is detection engineering: building the log pipelines, anomaly jobs and hunts that tell us when something on the network isn't right.

The AI side

A growing share of this work now happens alongside an AI coworker, and a lot of what I build is the plumbing that makes that possible.

  • Tool servers. MCP servers I wrote so an assistant can query the SIEM, read firewall state and work with other systems directly.
  • Skills. Repeatable playbooks for triage, enrichment and query building, so a good investigation method gets reused and not reinvented.
  • Local models. Running models on my own hardware, for work that shouldn't leave the building and to learn what small models can and can't do.
  • Memory. A notes vault with semantic search, so yesterday's context is still there tomorrow.

I write about what works, what breaks and what it costs.

What you'll find here

  • Hunt write-ups. How a real investigation went: what tipped us off, the queries, the dead ends and the verdict.
  • Build logs. The tools and pipelines behind the work, one problem at a time, with code.
  • AI-augmented engineering. What changes, and what doesn't, when an AI works the queue with you.
  • Short notes. Firewall, identity and SIEM gotchas in a few paragraphs.

Ground rules

Everything here comes from real work, and it gets cleaned up before it's published. No names, no addresses, nothing that points at a specific system. The numbers are real.

The opinions are mine and not my employer's.

Why "wandering"

Security work rarely goes in a straight line. You start on one alert and end up three systems away, having learned something you weren't looking for. These are the notes from those detours.