Principal Senior GPU SW Performance Engineer, Post-Training

Amd

Confirmed live yesterday High trust

Hybrid

Quick summary

Work type: Hybrid
Location: San Jose, CA
Salary: $240,000–$360,000 / yr
Posted: 170 days ago
Freshness: Confirmed live yesterday
Closes: Mar 25, 2027
Nearby: 99+ roles within 25 mi

Market check

Salary context

Above market

How this pay compares to similar roles

Similar $216k

This role $300k

$145k most similar roles pay here $383k

This role pays more than 92% of similar roles. Most pay $187,700–$244,000 — the shaded band above. At the midpoint, this role pays about $300k versus about $216k for comparable roles.

Based on 240 similar postings.

Employer

About Amd

AMD (Advanced Micro Devices) is a semiconductor company that develops high-performance processors, graphics cards, and adaptive computing solutions for gaming, data centers, and embedded markets. Industry: Semiconductors

Amd currently has 367 open roles on FindRole.

Listed pay typically runs $166,400–$249,600 across 367 roles with salary data.

Most-posted roles

View all roles at Amd

At a glance

TL;DR · Principal Senior GPU SW Performance Engineer, Post-Training

Role Posting Log in to save

As a Principal / Senior GPU SW Performance Engineer — Post‑Training, you will join the team to drive performance for post-training workloads on AMD Instinct GPUs. You will be responsible for delivering fast, stable, and reproducible training pipelines on ROCm by addressing cross-stack issues involving data loaders, kernels, distributed training, and compilers. Your daily work involves optimizing throughput, memory efficiency, and stability across data, model, and optimizer steps while improving multi-GPU/multi-node communication patterns. You will develop efficient kernels, perform graph-level optimizations, and resolve bottlenecks using standard profiling tools to prevent regressions in CI. The role requires expertise in PyTorch, Python, C++, Triton, and HIP for deep learning tasks like SFT, LoRA, and RL-based training at scale. You will collaborate with framework, compiler, and model teams to solve complex performance challenges within the AI domain.

Skills

ROCm HIP Triton PyTorch Python C++ Distributed Training GPU Performance Engineering SFT LoRA RL FSDP ZeRO multi‑GPU multi‑node

What you'll do

Drive performance for fine-tuning and RL training solutions on AMD GPUs.
Improve throughput, memory efficiency, and stability across data, model, and optimizer steps.
Optimize multi-GPU and multi-node training and communication patterns.
Develop efficient kernels, operations, and targeted graph-level optimizations.
Profile, diagnose, and resolve performance bottlenecks using standard tooling.
Prevent regressions in CI and ensure stable training pipelines on ROCm.
Deliver reproducible pipelines and documentation for internal and external developers.
Coordinate with framework, compiler, and model teams to implement durable improvements.

What we're looking for

Bachelor's, Master's, or Doctorate degree in Computer Science, Computer Engineering, Electrical Engineering, or a related field.
Proven experience in GPU performance engineering for deep learning using ROCm, HIP, Triton, or similar technologies.
Hands-on experience with SFT, LoRA, and RL-based training at scale.
Strong proficiency in PyTorch, including torch.distributed, FSDP, ZeRO, or equivalent frameworks.
Proficiency in Python and C++ with the ability to read and write kernels.
Experience with distributed systems and collective communication libraries.
Ability to profile, diagnose, and resolve performance bottlenecks while ensuring reproducibility.
Track record of turning profiles into fixes, upstreaming changes, and documenting results.