3 days ago
Remote, Germany +3 moreSenior
Responsibilities
- Characterize node, rack, and cluster performance using microbenchmarks and representative HPC and AI applications.
- Analyze Linux scheduling, NUMA, CPU and memory affinity, interrupt handling, I/O, networking, GPU utilization, and resource contention issues.
- Investigate distributed workload scaling, collective communication, synchronization delays, load imbalance, interconnect topology, and network traffic.
- Use profiles, traces, hardware counters, and cluster telemetry to isolate root causes and develop focused reproductions.
- Develop and maintain Python tools for benchmark automation, data collection, baseline creation, performance analysis, and regression detection.
- Define and execute performance validation for new platforms and cluster deployments, verifying that corrective actions restore workload performance.
- Collaborate across hardware, firmware, driver, library, and application teams to implement and verify fixes.
Requirements
- BS or MS in Computer Science, Engineering, Mathematics, or Physics, or equivalent experience.
- 8+ years of applicable experience.
- Extensive practical experience with Linux performance analysis, including processes, threads, CPU and memory architecture, NUMA, I/O, and networking.
- Experience investigating performance in multi-node HPC or GPU-accelerated systems.
- Deep expertise in at least one of GPU application performance, Linux systems performance, high-speed networking and collective communication, parallel storage, or HPC performance tooling.
- Proficiency in Python and experience developing and debugging performance-sensitive C, C++, and CUDA code.
- Ability to design controlled experiments, interpret profiling and counter data, and demonstrate reproducible performance improvements.
- Ability to independently drive technical efforts across hardware and software boundaries and communicate findings clearly.
- Preferred experience with Nsight Systems, Nsight Compute, Linux perf, eBPF-based tools, RDMA, MPI, NCCL, PCIe, NVLink, InfiniBand, Ethernet, Slurm, container runtimes, cgroups, distributed training or inference, scientific applications, strong- and weak-scaling studies, and ReFrame.
About Nvidia
Nvidia designs and sells GPUs and accelerated computing platforms for data centers, AI/ML, graphics, gaming, and automotive, monetizing through hardware, software platforms (CUDA, AI frameworks), and systems like DGX and networking. Customers include cloud providers, enterprises, researchers, and OEMs. Founded in 1993 and headquartered in Santa Clara, it is a public company traded on NASDAQ under NVDA.