Senior AI Infra Performance & Observability Engineer

CoreWeave; Sunnyvale (CA); Presencial; USD 182.000 - 242.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Medical insurance

Life Insurance

401(k) with match

Tuition Reimbursement

Flexible PTO

Descripción de la vacante

CoreWeave is seeking a Senior Engineer to drive Benchmarking & Performance, building observability systems that surface performance signals across data centers. You will turn billions of telemetry events into real-time insights guiding engineers, product teams, and executives.

The role emphasizes near real-time diagnostics for GPU fleets, interconnect fabrics, and distributed training/inference workloads, sitting close to hardware and workloads rather than dashboards.

Formación

• 5+ years in distributed systems, observability, or performance tooling for infra/ML
• Strong coding in Python or Go; familiarity with GPU infra
• Hands-on Kubernetes at production scale, CI/CD, observability stacks
• Knowledge of time-series DBs and PromQL/MetricsQL for alerting

Responsabilidades

• Design and build systems for continuous health and performance assessment of AI infrastructure
• Own time-series DB layer; write/optimize PromQL queries for alerts and trends
• Build pipelines for GPU telemetry, interconnect fabric health, and anomaly detection
• Design performance data lake with Iceberg/Parquet/Avro; support latency and throughput metrics
• Profile and tune query engines for low-latency observability; benchmark infra

Conocimientos

Distributed systems

Observability

Python

Go

Kubernetes

PromQL

Performance tooling

GPU infra

Communication

Herramientas

Prometheus

Grafana

OpenTelemetry

Descripción del empleo

CoreWeave is seeking a Senior Engineer to drive Benchmarking & Performance, building observability systems that surface performance signals across data centers. You will turn billions of telemetry events into real-time insights guiding engineers, product teams, and executives.

The role emphasizes near real-time diagnostics for GPU fleets, interconnect fabrics, and distributed training/inference workloads, sitting close to hardware and workloads rather than dashboards.