Principal Distributed Systems Software Engineer, Collectives - CA, San Jose, United States
Originally posted 28 October 2026 by the employer.
About the role
As a core member of the Collectives and Network Performance team, you will define and deliver AI pre-training and distributed inference performance on AMD GPUs. You will optimize collective communication libraries, scale-up and scale-out networking software, and distributed training and inference workloads across multi-GPU and multi-node systems.
What you'll do
- Define goals, scope, and roadmaps for collectives and network optimizations, aligning them with hardware, framework, and ROCm roadmaps. - Lead the design and development of collective communication algorithms and optimizations (e.g., all-reduce, all-to-all, all-gather) across scale-up and scale-out networks. - Lead performance tuning, profiling, and analysis of large-scale LLM, diffusion, multimodal, RecSys, and generative AI models across single-node and multi-node systems. - Participate in hardware-software co-design for scale-up networks, NICs, and scale-out networks, representing software performance requirements to hardware engineering teams. - Build frameworks, tools, and infrastructure for performance estimation, modeling, regression tracking, and reporting. - Collaborate with internal GPU library, compiler, and runtime teams, as well as open-source maintainers across PyTorch, JAX, vLLM, and SGLang.
What you'll need
- Master’s or PhD degree in Computer Science, Computer Engineering, Electrical Engineering, or a related field. - Advanced proficiency in C++ and technical expertise within Linux operating environments. - Knowledge of network, NIC, and GPU hardware architecture with hands-on experience mapping model architectures to low-level software and hardware. - Experience with performance engineering, communication libraries, networking, and GPU distributed systems.
Nice to have
- Deep expertise in designing and optimizing collective communication libraries such as RCCL, NCCL, or MPI, alongside scale-up, scale-out, and RDMA network architectures. - Experience integrating optimized communication routines into deep learning frameworks such as PyTorch, JAX, vLLM, and SGLang. - Experience with performance analysis, projection, and bottleneck identification across the full hardware and software stack. - Proficiency in Python along with experience in performance tuning, debugging, and robust test design for high-performance codebases.
Skills: ROCm, collectives, RCCL, NCCL, C++, MPI
This role has been open 0 days — well below the 55-day median for Infrastructure/Platform Software roles.
Infrastructure/Platform Software · GPU Software Stack
|
Open roles in category
258
|
Median days open
55 d
|
Median salary
$220k
|
See the full market breakdown ▾Category comparison, skills in demand, and who else is hiring
| Metric | AMD | All employers we track in this specialty (258 roles · 19 employers) |
|---|---|---|
| Open roles in this specialty | 55 | 258 |
| Open roles in the wider Software, Firmware & Systems family | 389 | 6395 · 147 employers |
| Median days open | 57 d | 55 d (+2 d vs this employer) |
| Median salary (USD postings) | — | $220k |
Skills observed across this category: ROCm, collectives, RCCL, NCCL, C++, MPI
Who's hiring in this category
- NVIDIA · 127 open roles · median 58 d
- AMD (this employer) · 55 open roles · median 57 d
- Intel Corporation · 29 open roles · median 15 d
- Qualcomm · 17 open roles · median 188 d
- Arm Holdings · 8 open roles · median 1 d
- Graphcore · 4 open roles · median 8 d
How we counted: 258 open Infrastructure/Platform Software (GPU Software Stack) roles from 19 employers tracked in the SemiconductorJobs index, counted 29 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.