Models in production
5
1 canary · 1 staging
Inferences (24h)
↑ 6.8%
2.41M
GPU utilization
73%
148/160 nodes
Drift alerts
2
TraumaVision · PulmoSeg
Drift monitoring · 30 days
KS distance vs. baseline distribution
threshold 2.0%
Model accuracy
Production fleet
96.8%
fleet avg
Inference throughput · 24h
Requests per hour across cluster
peak 3,842/h
Retraining pipelines
FractureNet v4.3.0ETA 12m
stage · Validation
BoneIQ Detect v2.6.0ETA 1h 42m
stage · Training
PulmoSeg v1.9.0ETA 3h 10m
stage · Data prep
TraumaVision v3.1.0ETA manual
stage · Approval queue
Explainability snapshots
Wrist FX heatmap90.0%
Pelvis ROI91.3%
Skull contour92.6%
Chest CT mask93.9%