Angel13 Systems is a four-person practice out of Bangkok. We are hired when a platform has outgrown the people who built it — when the network diagram is out of date, the build takes forty minutes, and the only person who understands the storage layer left in March.
How we work
We do not staff-aug. Each engagement is a bounded problem with a named owner, a two-week discovery, and a written runbook at the end. If your team cannot operate what we built after we leave, we did the job wrong.
Two weeks reading your code, your alerts, and your last six postmortems. You get a findings document whether or not you continue.
A single engineer holds context end to end. The other three review. No rotating bench, no handoff tax.
Deliverables are Markdown in your repo, Terraform in your state, and dashboards on your account. Nothing lives with us.
Scope is agreed before the first commit. Overruns are ours to absorb, which keeps our estimates honest.
Capabilities
We turn down work outside these areas. The list has not changed since 2021.
BGP, anycast, tunnel topologies, MTU archaeology, and the path-MTU bugs that only show up on one carrier.
Replication topology, backup restore drills that actually restore, and tiering that survives an audit.
Cache-correct pipelines, reproducible images, and rollback paths tested on a Tuesday rather than discovered on a Friday.
Datacentre exits, cloud moves, and version jumps with a cutover plan that names who calls it off and when.
Sample finding
Discovery output is blunt and reproducible. Every claim in a findings document ships with the command that produced it, so your team can re-run it after we are gone.
$ a13 probe --scope edge --since 30d WARN path-mtu blackhole on carrier B (v6 only) 1492 clamp missing on 3 of 14 tunnels ICMPv6 type 2 dropped upstream — invisible WARN build cache keyed on $CI_COMMIT_SHA hit rate 4.1% — effectively cold every run INFO restore drill last executed 2023-11-04 retention policy claims quarterly 3 findings · 2 blocking · report → findings.md
Send the symptom. If it is not our kind of problem, we will say so in the first reply and point you at someone better suited.
Start a scoping call