Principal Software Engineer, GPU Firmware and GPU System Software — CSP Engagements
NVIDIAfull time · Principal
$272,000 – $431,250
Listed on NVIDIA’s own careers site. You apply with them directly — we never stand between you and the employer.
What this role is
This role serves as the technical bridge between NVIDIA and hyperscale cloud providers, owning GPU firmware integration, fleet-scale deployment strategy, and the translation of customer operational challenges into product improvements. It suits someone with deep GPU architecture expertise and proven success managing mission-critical firmware updates across massive deployments.
Our summary, not NVIDIA’s wording. The full posting is on their site.
What they ask for
Required
- 15+ years GPU system software, firmware, or accelerator platform engineering experience
- BS or MS in Computer Science, Electrical Engineering, or related field (or equivalent)
- Deep understanding of GPU architecture internals: streaming multiprocessors, GEMM execution, compute kernels, memory hierarchy
- Understanding of multi-GPU fabric architectures like NVLink
- Understanding of GPU firmware architecture: VBIOS, microcontroller firmware, InfoROM, and driver stack interaction
- Experience with firmware update lifecycle management at scale: sequencing, A/B updates, rollback, staged rollout, recovery
- Understanding of GPU error handling and recovery flows
- Experience with GPU health monitoring and telemetry: Xid errors, thermal events, power events, ECC counters
- Customer obsession with proven success influencing engineering teams on quality and fleet manageability
Nice to have
- Direct experience with NVIDIA GPU VBIOS or microcontroller firmware or driver internals
- Background in GPU fleet management at 10K+ GPU scale
- Experience with GPU error taxonomy and building operational runbooks
- Understanding of GPU security: secure boot chain, code signing, attestation, debug authentication, multi-tenancy isolation
- Familiarity with GPU power management architecture and its impact on workload performance