K
MLOps Pipelines · Training & Deployment07:33:06Sun, May 24
AI Operations
Training Pipelines
Argo Workflows orchestrating de-identification, training, validation, canary, and promotion across the model fleet.
Active runs
3
Success rate (30d)
96.4%
Avg duration
2h 18m
GPU cost / week
$12.4k
Pipeline · run-8821 · NeuroVision v5.2.0-rc
Ingest
84,212 studies
De-ID
PHI scrubbed
Curate
labels QC
Train
8× A100 · 18/40
Validate
shadow holdout
Canary
5% rollout
Promote
production ring
Recent runs
| Run | Model | Trigger | Started | Duration | Step | Status |
|---|---|---|---|---|---|---|
| run-8821 | NeuroVision | auto·drift | 12:42 | — | Training (epoch 18/40) | running |
| run-8820 | FractureNet | scheduled | 08:00 | 2h 14m | Deployed to canary 5% | success |
| run-8819 | PulmoScan | manual | 06:21 | 3h 02m | Promoted to production | success |
| run-8818 | MammoAssist | auto·data | Yesterday 22:10 | 1h 48m | Validation gate: AUC dropped 1.2pp | failed |
| run-8817 | CardioMR | manual | Yesterday 18:00 | 4h 12m | Held in shadow ring | success |