ou own everything above the cluster: the observability platform, the GPU metering-to-billing pipeline, the operation of the AI service catalogue, and the SLOs that define reliability. Your job is to make the GPU clusters delivered to you observable, measurable, billable, and reliable.

What you will own-

• Observability & ReliabilityBuild and operate the full observability stack: metrics (Prometheus + long-term store such as VictoriaMetrics/Thanos), logs (Loki + a forwarder), and alerting (Alertmanager)
• Integrate DCGM GPU telemetry into the pipeline: utilisation, memory, temperature, power, SM activity — per GPU, per tenant
• Surface GPU and fabric health (DCGM XID errors, NCCL/RDMA, RoCE/link health) from the Infra and Network engineers into a single pane — you own dashboards, alerting, and tenant-impact SLOs; Infra/Network own the fabric remediation
• Build the dashboarding capability (Grafana with SSO), ready for tenants at launch
• Define SLOs, SLIs, and error budgets; implement burn-rate alerting
• Implement OpenTelemetry instrumentation across platform services; lay the alerting and runbook foundations for reliable operation later
•