Senior HPC Platform Architect
- Posted 18 hours ago
- Be among the first 10 applicants
Job Description
NVIDIA is looking for an exceptional engineer to grow and thrive alongside our HPC Infrastructure team that designs, evaluates, and optimizes the compute inrastructure powering NVIDIA's next-generation silicon design and AI
workloads. You will be the primary architecture reviewer and performance champion for new data center clusters being built across multiple sites globally.
What You'll Be Doing
workloads. You will be the primary architecture reviewer and performance champion for new data center clusters being built across multiple sites globally.
What You'll Be Doing
- Own data center architecture reviews for new HPC clusters — evaluating compute, storage,networking, and cooling decisions and challenging assumptions before clusters are built.
- Serve as the BDC representative in cluster build meetings across multiple simultaneous programs (GPU compute clusters, AI infrastructure, and EDA environments), driving architecture alignment independently.
- Analyze and validate cluster design choices across storage-to-compute distance, cross-mount latency, rack layout, and multi-site topology — and surface risk and tradeoff recommendations to leadership.
- Lead performance benchmarking and profiling of HPC cluster infrastructure, running sanity benchmarks, regression suites, and cluster health checks to identify bottlenecks early.
- Drive infrastructure optimization at multiple layers: scheduler-level tuning (LSF/Slurm),hardware-level tuning (GPU, CPU, networking, storage), and OS/kernel-level tuning (NUMA binding, socket binding, huge page configuration, kernel image selection).
- Collaborate with platform and operations teams on cluster health, capacity planning, and operational readiness for tapeout milestones.
- Partner with vendors and internal teams to evaluate new hardware, storage systems, and networking fabrics; produce architecture recommendations backed by data.
- Continuously improve infrastructure observability, benchmarking frameworks, and architecture documentation to raise the bar across the BDC HPC platform.
- B.E./B.Tech or M.Tech/M.S. with 5+ years of hands-on experience in HPC infrastructure, data center architecture, systems engineering, or a senior SRE/platform engineering role at scale.
- Deep understanding of data center architecture fundamentals: compute (CPU/GPU servers), storage (parallel file systems, NVMe, tiered storage), and high-speed networking (InfiniBand,Ethernet, NVLink).
- Proven ability to evaluate and challenge infrastructure design decisions — including rack layout, power/cooling constraints, storage-to-compute distance, and network fabric topology.
- Experience with OS and kernel-level performance tuning: NUMA binding, socket affinity, huge page configuration, kernel image selection, and system parameter optimization.
- Hands-on experience with large-scale Linux HPC cluster administration using workload managers such as LSF and/or Slurm, including job scheduling optimization and resource utilization analysis.
- Strong Linux/Unix system administration skills and proficiency in scripting (Python, Bash, or Perl) for automation and analysis.
- Experience running HPC performance benchmarks, cluster health checks, and profiling tools to identify infrastructure bottlenecks.
- Excellent problem-solving and communication skills — able to synthesize complex architectural tradeoffs and present clear recommendations to engineering leadership.
More Info
Key Skills
kernel image selection
HPC infrastructure
Linux Unix system administration
huge page configuration
OS and kernel-level performance tuning
NVMe tiered storage
cluster health checks
HPC performance benchmarks
storage parallel file systems
SRE platform engineering
NUMA binding
NVLink
compute CPU GPU servers
job scheduling optimization
Linux HPC cluster administration
high-speed networking
socket affinity
resource utilization analysis
About Company
Skills:
Data Center Architecture, Ethernet, NVLink, storage parallel file systems, high-speed networking, SRE platform engineering, compute CPU GPU servers, Profiling Tools, infiniband, OS and kernel-level performance tuning, resource utilization analysis, NVMe tiered storage, Systems engineering, HPC infrastructure, Linux HPC cluster administration, HPC performance benchmarks, cluster health checks, job scheduling optimization, NUMA binding, kernel image selection, Linux Unix system administration, huge page configuration, socket affinity