// CAREERS
Build the loop with us.
RunAgain is the self-improving loop for AI agents: trace what happened, evaluate what changed, replay it exactly as it ran, and turn every issue into a test. We are early, the surface area is enormous, and everyone here owns their work end to end.
Roles, levels and compensation are kept up to date there. Apply in one click.
What you would work on
OpenTelemetry pipelines that swallow every span, tool call and token from production agents, then make them searchable, diffable and replayable at volume.
Mocked environments, deterministic replays, LLM-as-judge and rubric scoring, dataset regression: the machinery that says whether a change actually made the agent better.
The app people live in all day: dashboards, trace views, the SDKs and the developer experience around them. Fast, dense, and pleasant to use.
Nothing open that fits?
Write anyway. Tell us what you would build and point at something you have shipped: tamas@runagain.ai. Good people get remembered when a role opens.