Lead GPU ML Training Performance Engineer

Advanced Micro Devices, Inc.; San Jose (CA); On-site; USD 120,000 - 160,000; Full time

Job summary

Advanced Micro Devices, Inc. is seeking a driven performance engineer to optimize posttraining workloads on AMD Instinct GPUs in San Jose, CA. The role involves collaborating on complex issues across various teams to enhance efficiency, stability, and throughput in training pipelines.

The ideal candidate will have a strong foundation in deep learning, proficiency in GPU performance engineering, and experience with tools like PyTorch. You will work on optimizing multiGPU training solutions and delivering reproducible pipelines for internal and external developers.

Qualifications

• Proven experience in GPU performance engineering for deep learning.
• Hands-on with SFT, LoRA and RL-based training at scale.
• Comfortable reading/writing kernels when needed.

Responsibilities

• Lead performance for finetuning and RL training solutions on AMD GPUs.
• Improve throughput, memory efficiency, and stability across data, model and optimizer steps.
• Optimize multiGPU/multinode training and communication patterns.

Skills

GPU performance engineering

Deep learning frameworks (ROCm/HIP, Triton)

Strong PyTorch experience

Proficient in Python and C++

Experience with distributed systems

Education

B.S./M.S./Ph.D. in Computer Science, Computer Engineering, or Electrical Engineering

Job description

Advanced Micro Devices, Inc. is seeking a driven performance engineer to optimize posttraining workloads on AMD Instinct GPUs in San Jose, CA. The role involves collaborating on complex issues across various teams to enhance efficiency, stability, and throughput in training pipelines.

The ideal candidate will have a strong foundation in deep learning, proficiency in GPU performance engineering, and experience with tools like PyTorch. You will work on optimizing multiGPU training solutions and delivering reproducible pipelines for internal and external developers.