> Markdown version of [/jobs/ext/4775081-principal-gpu-software-performance-engineer-posttraining](https://www.wearedevelopers.com/jobs/ext/4775081-principal-gpu-software-performance-engineer-posttraining). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Principal GPU Software Performance Engineer - PostTraining - **Company:** Advanced Micro Devices, Inc. - **Location:** San Jose, CA, United States - **Salary:** $240,000 - **Contract:** Permanent contract - **Skills:** Artificial Intelligence, Computer Engineering, Graphics Processing Unit (GPU), Deep Learning, Information Technology, ROCm, Software Performance - **Published:** November 18, 2026 - **Apply:** https://jobs.localjobnetwork.com/apply/add/86678264/1 ## About the Role * ProvenGPU performance engineeringfor deep learning (ROCm/HIP, Triton, or similar). * Handson withSFT. LoRAandRLbased trainingat scale. * StrongPyTorchexperience (torch.distributed, FSDP/ZeRO or equivalent). * Proficient inPythonandC++; comfortable reading/writing kernels when needed. * Experience withdistributed systemsand collective communication libraries. * Track record ofturning profiles into fixes, upstreaming changes, and documenting results. ACADEMIC CREDENTIALS: * B.S./M.S./Ph.D. in Computer Science, Computer Engineering, Electrical Engineering, or equivalent ## Description Drive the performance ofposttrainingworkloads on AMD Instinct GPUs. You'll work across kernels, distributed training, and framework integrations to deliver fast, stable, and reproducible training pipelines on ROCm., * Lead performance forfinetuning and RLtraining solutions on AMD GPUs. * Improvethroughput, memory efficiency, and stabilityacross data, model, and optimizer steps. * OptimizemultiGPU/multinodetraining and communication patterns. * Contributeefficient kernels/opsand targeted graphlevel optimizations. * Profile, diagnose, and resolvebottlenecks using standard tooling; prevent regressions in CI. * Shipreproducible pipelines and documentationadopted by internal teams and external developers. * Collaborate withframework, compiler, and modelteams to land durable improvements., AMD may use Artificial Intelligence to help screen, assess or select applicants for this position. AMD's "Responsible AI Policy" is available here.