AI Systems Performance Engineer - USA-CA San Jose Innovation Drive, United States
$141,300 - $226,000 USD yearly
Originally posted 20 April 2026 by the employer.
Benchmark AI inference and training workloads on GPU clusters, focusing on NIC, driver, and RDMA/RoCEv2 performance.
About the role
This role benchmarks AI inference, training, and storage workloads on GPU clusters within the Architecture group's Performance Lab, focusing on NIC and host networking performance. The primary focus is the server side of the network, including the NIC, drivers, and the host stack connecting AI workloads to the fabric.
What you'll do
- Install, configure, and run AI performance benchmarks, including MLPerf (Training and Inference) and NCCL collective communication tests. - Optimize NIC, driver, and host network settings, including RDMA/RoCEv2, congestion control, queue and interrupt configuration, and NUMA placement for distributed AI workloads. - Isolate performance bottlenecks across the Linux OS, server hardware, NICs, and the network. - Build tools and frameworks to ensure repeatable benchmarking. - Document test methodologies and present findings and recommendations to hardware, software, and networking teams.
What you'll need
- Bachelor’s or Master’s degree in Computer Science, Computer Engineering, Electrical Engineering, or a related field, typically with 12+ years of related industry experience. - Strong hands-on Linux experience, including system-level performance tuning and troubleshooting. - Proficiency in Python or a similar scripting language. - Deep experience in performance tuning or validation of high-speed NICs, drivers, or RDMA/RoCEv2, or bringing up and tuning multi-node GPU clusters with collective communication libraries like NCCL. - Proven ability to diagnose root causes of performance problems in distributed systems. - Authorization to work in the US.
Nice to have
- Experience with MLPerf Training or Inference. - Familiarity with Ethernet switch and fabric behavior, including PFC, ECN, and load balancing. - Experience with Docker, Kubernetes, or Slurm in AI deployments. - Experience building CI/CD pipelines for performance regression testing.
Skills: AI inference, MLPerf, Ethernet fabric, AI workloads, hardware performance, system bottlenecks
This role has been open 187 days — well above the 87-day median for AI/ML Hardware Engineering roles.
AI/ML Hardware Engineering · AI ML Hardware
|
Open roles in category
506
|
Median days open
87 d
|
Median salary
$236k
|
See the full market breakdown ▾Category comparison, skills in demand, and who else is hiring
| Metric | Broadcom | All employers we track in this specialty (506 roles · 77 employers) |
|---|---|---|
| Open roles in this specialty | 3 | 506 |
| Open roles in the wider Software, Firmware & Systems family | 80 | 6250 · 144 employers |
| Median days open | 30 d | 87 d (−57 d vs this employer) |
| Median salary (USD postings) | — | $236k |
Skills observed across this category: AI inference, MLPerf, Ethernet fabric, AI workloads, hardware performance, system bottlenecks
Who's hiring in this category
- Qualcomm · 105 open roles · median 116 d
- NVIDIA · 89 open roles · median 58 d
- AMD · 33 open roles · median 107 d
- Micron Technology · 26 open roles · median 59 d
- Mobileye · 20 open roles · median 85 d
- Analog Devices · 18 open roles · median 43 d
How we counted: 506 open AI/ML Hardware Engineering (AI ML Hardware) roles from 77 employers tracked in the SemiconductorJobs index, counted 25 Oct 2026. Specialty figures count only roles carrying this exact specialty label, so an employer's related work in neighbouring specialties is not included there — it is counted in the wider Software, Firmware & Systems family row. Figures refresh nightly.