The Principal/Senior GPU Software Performance Engineer role at AMD focuses on enhancing the performance of post-training workloads on AMD Instinct™ GPUs. The position involves collaboration across various teams to optimize training pipelines, ensuring they are fast, stable, and reproducible. The ideal candidate will lead complex issues related to data loaders, kernels, and distributed training, driving measurable improvements. This role is crucial for advancing AMD's mission in AI and computing.

• Proven experience in GPU performance engineering for deep learning, specifically with ROCm/HIP, Triton, or similar technologies.
• Hands-on experience with SFT, LoRA, and RL-based training at scale.
• Strong experience with PyTorch, including torch.distributed and FSDP/ZeRO or equivalent.
• Proficiency in Python and C++, with the ability to read and write kernels as needed.
• Experience with distributed systems and collective communication libraries.
• A track record of turning profiles into fixes, upstreaming changes, and documenting results.
• B.S./M.S./Ph.D. in Computer Science, Computer Engineering, Electrical Engineering, or equivalent.

• Lead performance optimization for finetuning and reinforcement learning 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.
• Contribute efficient kernels and targeted graph-level optimizations.
• Profile, diagnose, and resolve bottlenecks using standard tooling, while preventing regressions in continuous integration.
• Ship reproducible pipelines and documentation that are adopted by internal teams and external developers.
• Collaborate with framework, compiler, and model teams to implement durable improvements.

• AMD provides a competitive 'Total Rewards' package that focuses on financial growth, health, and work-life balance.