Senior System Software Engineer, Performance - CUDA Driver

We are looking for senior systems software engineers to make the CUDA Driver faster, more efficient, and ready for the next generation of accelerated computing. Our team develops performance-critical CUDA Driver features and systems software improvements that help AI, deep learning, HPC, and other CUDA-powered applications realize more of the performance available from NVIDIA GPUs!

Some of the hardest performance problems emerge not within one component, but at the boundaries among systems software, CPUs, interconnects, and GPUs. In this role, you will trace those problems from real workloads through the software and hardware stack, design and ship production features, and deliver validated optimizations for current and emerging platforms. 

You will combine hands-on engineering with broad technical influence, collaborating across CUDA software, hardware architecture, frameworks, applications, product, and customer-facing teams. You will lead cross-layer investigations, mentor engineers, and help set performance direction. Your work will help developers get more useful computing from NVIDIA GPUs today while helping shape the CUDA software and GPU architectures NVIDIA builds next. 

What you'll be doing: 

• Develop and ship performance-centric CUDA Driver features and programming-model capabilities from design through validation. 

• Diagnose complex performance problems through workload analysis, focused measurement and modeling, cross-layer root-cause isolation, and application-level validation. 

• Optimize critical CUDA primitives, memory management and movement, CPU–GPU coordination, and interconnect paths for latency, throughput, bandwidth, efficiency, and scalability. 

• Establish performance goals for current and future platforms, characterize as new platforms come online, close software and hardware gaps, and drive performance readiness through releases. 

• Translate workload and platform evidence into CUDA API, programming model, system software, and future GPU architecture recommendations. 

• Set subsystem performance direction and mentor engineers tackling complex systems and performance challenges. 

• Influence technical decisions with clear performance evidence and tradeoffs and strengthen implementations through rigorous design and code reviews. 

What we need to see: 

• A BS, MS, or PhD in Computer Science, Computer Engineering, Electrical Engineering, or a related field—or equivalent practical experience — with at least 7+ years of relevant systems-software development experience. 

• Strong production C/C++ systems-programming experience, including delivery of substantial features, optimizations, or production fixes in a complex codebase. 

• Strong operating systems and concurrency foundations, including threads, synchronization, processes, virtual memory, and user/kernel interactions. 

• Strong computer architecture foundations, including processors, memory hierarchy, caching and coherence, data movement, and system interconnects. 

• Demonstrated success improving real software performance: measuring behavior, identifying bottlenecks, implementing optimizations, and profiling to prove their effectiveness. 

• Sound technical judgment, ownership of ambiguous problems, and clear communication across organizational and disciplinary boundaries. 

• Direct CUDA or GPU experience is valuable but is not required when accompanied by deep systems software, operating systems, computer architecture, and performance-engineering foundations. 

Ways to stand out from the crowd: 

• Experience developing GPU or accelerator drivers, runtimes, kernel software, firmware, compilers, or other performance-critical low-level systems. 

• Experience with pre-silicon analysis, platform bring-up, performance modeling, or hardware/software co-design. 

• Systems-level performance experience with AI/DL, HPC, graphics, automotive, robotics, or similarly demanding workloads. 

• Evidence of technical inventions such as software-performance patents, novel production designs, or measurement-backed recommendations that influenced hardware revision or future architecture. 

• Python or another scripting language used for focused experimentation, data analysis, or visualization.