piqc scans your Kubernetes cluster (Read-only) and identifies which models are running on the wrong GPU tier and what the cost attribution is. It runs in a minute. I'd like to hear the community's experiences/thoughts on our detection approach and its benefits.