Control Plane

acme.health · production workspace · us-east-2 primary

GPU Hours (24h)
12,847
+8.2%· vs. yesterday
Active Endpoints
184
+6· 3 canary
Inference RPS
42.6k
+12.4%· p50 34ms
Training Jobs
76
-3· 12 queued
Datasets
1,204
+42· 83 versions
Enclave Uptime
99.998%
SLA· TEE attested

Compute throughput

Inference RPS · Training tokens/s · Federated bandwidth — last 24h

Inference RPSTraining tokens/sFederated bandwidth

Inference latency

p50 34ms · p99 128ms · SLO 250ms

Cluster health

us-east-2 · GPU pool A
71°C
106/128GPUs
83% utilized22 idle
us-east-2 · GPU pool B
58°C
22/64GPUs
34% utilized42 idle
eu-west-1 · GPU pool
74°C
41/48GPUs
85% utilized7 idle
us-west-2 · Confidential
62°C
18/32GPUs
56% utilized14 idle

GPU allocation

by workload class

Fine-tuning42%
Inference34%
Federated12%
Notebooks8%

Compliance posture

Live
  • HIPAA controls94 / 96
  • SOC 2 Type IIVerified
  • GDPR data map3 gaps
  • ISO 27001In review
  • PCI-DSS scopeOut of scope
Enclave attestation valid
PCR7: a1f4…9c02 · expires in 42d

Active training jobs

Live queue · 6 running · 12 queued

job_a7c1cortex-med-7b — SFT round 12
8 × H100· us-east-2· ETA 1h 12m· $284.10
68%Running
job_b920radiology-vit — LoRA sweep
4 × A100· eu-west-1· ETA 3h 04m· $96.80
42%Running
job_c412fraud-detector — retrain
2 × A10· us-west-2· ETA -· $0.00
0%Queued
job_d001voice-triage — RLHF
16 × H100· us-east-2· ETA 18m· $1,204.22
91%Running

Activity

Audit stream · last 30m

  • james.okafor@acme.health promoted model cortex-med-7b v4.2.1 → Production
    10:42:11 · 10.0.4.12
  • system attested enclave med-inference-tee-01
    10:38:02 · -
  • priya.mehta@acme.fin rotated key kms/prod/fraud-detector
    10:31:44 · 10.0.3.51
  • anna.weiss@acme.health started job radiology-vit — LoRA sweep
    10:22:09 · 10.0.4.88
  • sara.kim@acme.health deployed canary voice-triage 10%
    10:04:52 · 10.0.4.31
  • system federated round committed readmit-federated round 42
    09:52:11 · -

Inference endpoints

Sorted by traffic

EndpointRegionRPSp50p99ErrStatus
cortex-med-7bTEE
ep_prod_med · 6 / 12
us-east-22.4k34ms128ms0.02%Healthy
fraud-detector
ep_prod_fraud · 18 / 24
multi-region12.8k8ms22ms0.00%Healthy
radiology-vitTEE
ep_prod_rad · 3 / 6
eu-west-141062ms184ms0.14%Degraded
kyc-classifierTEE
ep_prod_kyc · 4 / 8
us-west-21.1k12ms48ms0.01%Healthy
voice-triage (canary 10%)TEE
ep_canary_voice · 2 / 4
us-east-2180220ms610ms0.44%Canary
risk-scorer
ep_prod_risk · 9 / 12
multi-region3.6k6ms18ms0.00%Healthy

Featured model

cortex-med-7b · v4.2.1

94.1%
Eval accuracy
Base
Llama-3.1-8B
Params
7.4B
Framework
PyTorch
Owner
James Okafor