Capabilities · Classification
Fraud Detection
HealthyIn-context predictionEndpoint fraud-detectionProduction v7Challenger v8
Fraud Detection / Deployments
Simulated
BLUE— Production
Healthyv7
- Traffic
- 100%
- AUC
- .947
- P95
- 43ms
- Cost / 1K
- $0.18
- Engine
- Managed A
- Created
- Last week
GREEN— Candidate
Readyv8
- Traffic
- 0%
- AUC
- .956
- P95
- 39ms
- Cost / 1K
- $0.16
- Engine
- Managed A · new model release, richer context
- Created
- Yesterday
Live traffic
BLUE v7100%
GREEN v80%
Deploy v8
Choose how the challenger reaches production
Production
v7
Challenger
v8
Customer responses still come from v7. v8 predictions are evaluated silently on 10% of mirrored requests.
No customer impact. Shadow traffic is billed at challenger cost.
Version timeline
Today
- 09:42Shadow evaluation completed — v8 beat v7 on 5/5 metrics
Yesterday
- 18:04v8 created — new foundation model release beat v7 on held-out examples (+.009 AUC)
Monday
- 11:21Feature drift detected on merchant_category — severity low
Last week
- 16:17v7 promoted to production — blue-green completed