I work on the parts of a network that decide whether everything else
behaves: how routes are chosen, how traffic is segmented and carried,
and where the authoritative record of it all actually lives.
Most network problems I care about are really
data problems. A single source
of truth turns a network from a collection of device configs into something
you can query, validate and automate against — so changes stop being folklore
and start being reproducible.
When something breaks, I'd rather read the packets than guess. Capture
first, form a theory second — Wireshark has settled more arguments than
any diagram.
Lately I'm AI curious: poking
at where models are genuinely useful in network operations, and where they
are just an expensive way to be confidently wrong.