AI Platform Engineer

About the role

An organization can use SalexHub's AI credits or, on the plans that allow it, its own provider key. It can set limits for each agent, workflow and API key, and let the Agent work in the background on a schedule. All of that has to be exact, cheap to run and easy to see into.

You make sure a model call goes where the organization chose, costs what we said it would, and that a long job survives a restart.

What you would do

• Route each model call by the organization's own choices: our credits or its key, the models it allows and the limits it set.
• Meter every call and every run in AI credits, so that what a customer sees is what happened.
• Run background work on the queue: scheduled runs of the Agent and of workflows, with retries that are safe to repeat and budgets that stop a run that costs too much.
• Make model calls observable: latency, errors, refusals and cost for each organization, agent and workflow, in one place.
• Design and test the controls that keep data a model must not see out of what it reads.
• Add providers and models safely, and say plainly what changes when one is withdrawn.

What we look for

• Experience running queues or background jobs in production, with retries, idempotency and limits.
• Strong TypeScript and PostgreSQL.
• You have worked with more than one model provider, and you know that they differ in limits, errors and price.
• You have built metering or billing where the numbers had to be exact.
• You like to make a system easy to see into before you make it clever.