What Trace finds
Idle spend, failed jobs, and drift — in plain English.
Idle GPUs and cost
GPUs that sit running with no work on them, and what that costs you.
Failed jobs
Why your last job stopped, for example running out of GPU memory.
Staging vs production
Settings that quietly differ between environments.
What changed in the last deploy
The change that started the problem, tied back to the deploy.
Who it is for
ML engineers
You train and serve models on rented GPUs and need a clear cause when a run dies.
GPU cloud users
You keep pods and studios warm and want idle spend and failed jobs caught early.
Small teams without a platform hire
You share GitHub and one cloud account and need one place to ask what broke.