Description

What you'll be doing:

Lead the implementation of performance practices in large-scale GPU infrastructure, delivering powerful tools, methodologies, and flows to validate and improve multiple datacenter products concurrently.

Align next-generation AI workloads with next-generation datacenter builds for NVIDIA GPUs, CPUs, and networking hardware. Engage early with HW/FW/SW/platform internal and customer teams.

Develop engineering solutions that provide continuous insights into the performance of AI workloads in evolving environments, generating swift insights into improvements and regressions.

Decompose high-complexity performance or stability issues into minimal reproduction cases, working towards identifying the root cause.

Participate in collaborations with various SW and FW teams (BMC/SBIOS/OS/drivers, etc.) to develop outstanding methods and tools. Analyze, debug, and resolve critical firmware and software issues to achieve the highest AI workload performance at scale.

What we need to see:

Proven understanding of accelerated computing software stacks (CUDA).

Experience with modern cloud and container-based enterprise computing architectures, with Slurm preferred.

Strong programming and scripting experience in C/C++/Python/Bash.

Deep expertise in systems architecture and the impact of various components on performance.

Experience with container technology and Linux-based OSes, with Docker preferred.

Experience supporting high-performance computing or deep learning in engineering or academic research communities.

Strong teamwork and communication skills, coupled with results-focused analytical abilities.

BS in Engineering, Mathematics, Physics, or Computer Science (or equivalent experience); MS or PhD desirable with 8+ years of applicable experience

Key Skills

Education

Any Graduate