AMD is seeking a Principal-level PyTorch training framework expert to enhance the performance, scalability, and correctness of large-scale AI training on AMD Instinct™ accelerators. This role involves collaboration with various teams, including compiler, kernel, driver, and architecture, to optimize PyTorch internals and distributed training. The ideal candidate will have strong technical leadership skills and a hands-on approach to solving complex systems problems. This position is critical in driving industry-leading training performance and improving developer experience.

• Deep experience with PyTorch internals and training systems.
• Strong distributed training expertise, including DDP and FSDP.
• Proven track record in performance engineering and memory optimization.
• Strong programming skills in Python and C/C++.
• Familiarity with PyTorch ecosystem components and performance tooling.
• Experience working in Linux-based environments.

• Act as a technical authority for PyTorch training at AMD.
• Drive optimization of key PyTorch training workloads across systems.
• Improve and debug training performance in various areas.
• Partner with teams to resolve performance bottlenecks and correctness issues.
• Contribute to and influence upstream PyTorch discussions and code contributions.
• Develop and maintain training benchmarks and performance regression detection.
• Lead investigations of performance regressions and correctness issues.
• Mentor engineers and enhance framework-quality code and practices.
• Engage with customers on training enablement and best practices.

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