Senior AI Performance Architect: GPU-Scale Training Lead

AMD; Austin (TX); On-site; USD 180,000 - 280,000; Full time

Job summary

AMD is seeking a Senior Engineer to lead the next generation of AI infrastructure, defining GPU architecture specs to enable massive model training at scale. You will push 2-3x performance gains in training and inference through innovative system design and optimization.

You will champion advanced parallelization strategies, memory-efficient training, and collaboration with internal teams to implement state-of-the-art models and techniques across AMD’s AI stack.

Qualifications

• Extensive experience optimizing large-scale ML systems and GPU architectures.
• Strong CUDA programming skills and knowledge of memory hierarchies.
• Proven ability to design distributed training systems at scale.

Responsibilities

• Lead performance modeling and optimization for multi-trillion parameter LLM training/inference (text, vision, speech).
• Model and optimize parallelization across tensor, pipeline, context, expert and data dimensions.
• Architect memory-efficient training systems with pruning, quantization, and efficient batching.
• Extend SOTA models (e.g., GPT-4 family) and multi-modal architectures.
• Collaborate with researchers and stakeholders to disseminate results and iterate rapidly.

Skills

CUDA programming

GPU architectures

Distributed ML

Model parallelism

Transformer models

Performance optimization

Leadership

Communication

Education

Bachelor's/Master/PhD in CS/Engineering

Tools

PyTorch

TensorRT

OpenAI Triton

Megatron-LM

NSight Compute

nvprof

Ray

InfiniBand

Job description

AMD is seeking a Senior Engineer to lead the next generation of AI infrastructure, defining GPU architecture specs to enable massive model training at scale. You will push 2-3x performance gains in training and inference through innovative system design and optimization.

You will champion advanced parallelization strategies, memory-efficient training, and collaboration with internal teams to implement state-of-the-art models and techniques across AMD’s AI stack.