Station supervisor

The supervisor is the top tier of a three-tier automation platform driving through-hole placement machines, adapting a decision-worker-client pattern already proven on a sister team's line. It receives messages from the line, decides what logic the situation calls for, and instructs the worker below it. Machine-originated messages bubble back up the same path.

One supervisor coordinates one line, not one machine. Six are in production covering seven machines, because a line can hold more than one. That is the unit the decision logic is written against: what the line should do next is a question about the line, and a supervisor that owned a single machine could not answer it.

The pattern's separation of decision-making from machine communication means the logic can be reasoned about and changed without touching protocol code, and a single supervisor can coordinate a station whose lower tiers know nothing about why they are being asked to do something.

I implemented all three tiers for through-hole automation, then containerised and deployed the stack on Kubernetes through a sister team's monorepo. This platform was its first consumer. I drove the containerisation and adoption rather than building the deployment platform itself.