GPU Cloud Infrastructure at CoreWeave
Running large model training jobs across tens of thousands of GPUs requires strict hardware reliability. CoreWeave deploys custom telemetry agents directly on bare-metal host nodes, interrogating InfiniBand adapters and PCIe bus topologies with low latency. Node issues trigger instant rescheduling. Our control plane leverages eBPF probes and Rust system daemons to intercept packet drops and RoCE congestion signals before training runs fail. Host operating systems bypass standard virtualization layers to deliver raw hardware performance. This setup guarantees maximum compute density with predictable networking characteristics.
As a Systems Infrastructure Engineer, you design and harden the low-level platform services supporting frontier AI foundation labs. Disclosed compensation ranges from $182,000 to $242,000 USD base salary.
Core Production Responsibilities
- Build resilient systems daemons in Rust and Go to manage bare-metal compute nodes and GPU server sleds.
- Monitor high-speed InfiniBand switches, RoCE network fabrics, and NVMe-oF storage controllers.
- Develop eBPF kernel tracing tools to diagnose network packet degradation and driver hang conditions.
- Coordinate automated node validation workflows to isolate degraded hardware before tenant workload assignment.
Technical Background
- Demonstrated experience in Linux systems administration, low-level networking, and systems programming.
- Working knowledge of Linux kernel internals, device drivers, PCIe architectures, or container runtimes.
- Hands-on familiarity with scripting and systems languages including Rust, Go, or Python.
- Compensation: $182,000 to $242,000 USD base salary alongside equity packages and robust benefits.