System Software Engineer – Data Center GPU Compute Diagnostics

NVIDIA•Durham, NC

2d•$124,000 - $195,500

About The Position

NVIDIA is seeking a system software engineer to develop low-level diagnostic software for next-generation data center GPUs and rack-scale AI systems. The team builds software to exercise and validate complex hardware, including compute engines, memory and cache subsystems, NICs, PCIe and NVLink interfaces, power delivery, and thermal behavior. The position is suited to engineers with experience in embedded software, firmware, device drivers, hardware validation, or systems software. Relevant experience may come from GPUs, CPUs, networking, storage, servers, embedded systems, or other complex silicon-based products. Prior GPU, CUDA, or GEMM experience is helpful but not required; the engineer can learn GPU architecture and programming while working with experienced engineers. The engineer will own well-scoped diagnostic software components from design through implementation, validation, productization, and field support. The role collaborates with hardware architects, driver developers, silicon-validation engineers, manufacturing teams, and field engineers to bring up new hardware and diagnose difficult system failures. Diagnostic and stress workloads range from low-level GPU hardware tests to higher-level AI workloads, with opportunities to develop expertise in CUDA, GEMM-style compute, NCCL, and PyTorch-based workloads.

Requirements

BS or MS degree in Electrical Engineering, Computer Engineering, Computer Science, or a related field, or equivalent experience.
At least 2 years of experience in embedded software, firmware, Linux device drivers, systems software, hardware validation, diagnostics, or silicon bring-up.
Strong programming skills in C and C, plus working proficiency in Python.
Experience developing software that interacts with hardware, firmware, device drivers, hardware registers, or low-level interfaces.
Experience creating diagnostics, validation tests, stress tests, manufacturing tests, or other software used to isolate hardware or system failures.
Understanding of fundamental computer architecture concepts such as memory, caches, interrupts, DMA, buses, and device I/O.
Strong debugging and problem-solving skills, including the ability to investigate failures across hardware and software boundaries.
Ability to take ownership of a well-scoped problem and drive it to completion while collaborating with a technical lead and cross-functional teams.
Good written and verbal communication skills.

Nice To Haves

Prior GPU, CUDA, or GEMM experience.
Interest in developing expertise in CUDA programming, GEMM-style compute, NCCL, and PyTorch-based workloads.

Responsibilities

Develop diagnostic and stress software in C/C++ and Python for complex hardware systems.
Interface with hardware blocks, firmware, Linux device drivers, registers, telemetry, and low-level debugging tools.
Bring up and validate new silicon and system features using pre-production hardware and software.
Create targeted tests for compute engines, memory and cache subsystems, DMA engines, PCIe/NVLink interfaces, power, and thermal behavior.
Investigate hardware and software failures involving memory errors, ECC, data integrity, performance, thermals, voltage/frequency behavior, and high-speed interfaces.
Contribute to diagnostic and stress workloads ranging from low-level GPU hardware tests to higher-level AI workloads.
Use modern development and analysis tools, including AI-assisted tools where appropriate, to accelerate coding, debugging, test creation, and failure analysis.
Own well-scoped diagnostic software components from design through implementation, validation, productization, and field support.
Collaborate with hardware architects, driver developers, silicon-validation engineers, manufacturing teams, and field engineers.