Senior AI GPU Cluster Performance Engineer

Socket.dev; Austin (TX); Hybrid; USD 120,000 - 160,000; Full time

Benefits offered by this job

AMD benefits at a glance

Job summary

AMD in Austin seeks a GPU Cluster Network Performance Attainment Engineer to optimize and attain peak performance for GPU clusters, focusing on RDMA networks used in AI Clusters. You will work on data flows between GPU, NIC and cluster network and shape long-term strategy.

The role requires hands-on problem-solving, debugging HW/FW and drivers, and staying ahead of industry trends to provide strategic insights to senior management. Hybrid work is expected.

Qualifications

• Strong background in GPU architectures and parallel computing.
• Experience with RDMA networks and NIC/GPU data flows.
• Scripting proficiency (Python, Bash) for automation and performance analysis.
• Ability to debug HW/FW and drivers.

Responsibilities

• Scalability testing of GPU clusters under various workloads and configurations (RoCE & IB).
• Performance profiling to identify bottlenecks and provide actionable insights.
• Performance tuning including protocol enhancements and load-balancing.
• Documentation of performance analyses, tuning efforts, and outcomes.
• Collaboration with hardware, software, and system architects to integrate improvements.
• NIC & performance optimization focusing on RDMA throughput and latency.
• Benchmarking and analysis of baseline performance and bottlenecks.
• Continuous learning of GPU architectures and HPC trends.

Skills

GPU architectures

Parallel computing

System-level performance tuning

Debug methodologies

RDMA networking

Scripting: Python

Linux kernel networking

Performance profiling

Performance optimization

RDMA drivers

Education

Bachelors or Masters in electrical or computer engineering

Tools

Profiling tools

Job description

AMD in Austin seeks a GPU Cluster Network Performance Attainment Engineer to optimize and attain peak performance for GPU clusters, focusing on RDMA networks used in AI Clusters. You will work on data flows between GPU, NIC and cluster network and shape long-term strategy.

The role requires hands-on problem-solving, debugging HW/FW and drivers, and staying ahead of industry trends to provide strategic insights to senior management. Hybrid work is expected.