Senior GPU Kernel Architect for AI & ML

Advanced Micro Devices, Inc.; San Jose (CA); On-site; USD 190,000 - 250,000; Full time

Job summary

Advanced Micro Devices, Inc. (AMD) seeks a senior technical contributor to design, develop, and optimize high-performance GPU kernels for LLM and ML workloads.

You will analyze bottlenecks and craft architecture-aware optimizations to boost compute efficiency and memory bandwidth on AMD GPUs. The role involves close collaboration with GPU architecture, compiler, library, and framework teams to deliver scalable solutions across current and next-generation platforms.

Qualifications

• Strong C++ and parallel programming skills.
• Deep understanding of GPU compute, memory hierarchy, and bottlenecks.
• Experience developing and optimizing GPU kernels with HIP/CUDA or similar.
• Analytical mindset with attention to detail; able to work independently.

Responsibilities

• Architect and optimize AI workloads, including LLMs and quantized models for AMD GPUs.
• Design, implement, profile, and optimize GPU kernels for high hardware utilization.
• Conduct deep performance characterization across compute, memory, caching, and synchronization.
• Advance compiler/code-generation techniques using LLVM/MLIR for portable GPU performance.
• Explore AI-assisted development tools for code generation, debugging, and profiling.

Skills

C++ programming

GPU compute

Performance analysis

CUDA

HIP

LLVM/MLIR

Analytical/problem solving

Education

Bachelor's or Master's in electrical or computer engineering

Tools

CUDA

HIP

C/C++

Job description

Advanced Micro Devices, Inc. (AMD) seeks a senior technical contributor to design, develop, and optimize high-performance GPU kernels for LLM and ML workloads.

You will analyze bottlenecks and craft architecture-aware optimizations to boost compute efficiency and memory bandwidth on AMD GPUs. The role involves close collaboration with GPU architecture, compiler, library, and framework teams to deliver scalable solutions across current and next-generation platforms.