Senior GPU Compiler & Performance Engineer

AMD; San Jose (CA); On-site; USD 180,000 - 300,000; Full time

Job summary

AMD in San Jose, CA seeks a Senior Staff Software Engineer to design and implement compiler optimizations for GPU architectures using LLVM/MLIR. You will develop and optimize GPU kernels with CUDA or HIP, analyze performance bottlenecks, and collaborate across teams to enable new hardware features.

You will build profiling tools and contribute to code generation and kernel scheduling to maximize hardware utilization.

Qualifications

• Experience with ML workloads, AI frameworks, or HPC systems is required.
• Familiarity with GPU kernel generation, memory hierarchy optimization, and low-level performance tuning.
• Experience across hardware/software co-design and influencing architecture decisions.
• Strong understanding of GPU architecture and parallel computing concepts.
• Proficiency in C/C++ with experience in CUDA, HIP, or equivalent GPU programming models.
• Hands-on experience with compiler technologies such as LLVM, MLIR, or similar.
• Experience in performance analysis, profiling, and optimization of compute-intensive workloads.
• Solid understanding of compiler design, optimization techniques, and code generation.

Responsibilities

• Design and implement compiler optimizations using LLVM and/or MLIR targeting GPU architectures.
• Develop and optimize GPU kernels using CUDA, HIP, or similar parallel programming models.
• Analyze application performance, identify bottlenecks, and deliver end-to-end performance improvements across the stack.
• Collaborate with architecture, runtime, and framework teams to enable new hardware features and improve efficiency.
• Build tooling and methodologies for profiling, benchmarking, and performance analysis on GPU systems.
• Contribute to compiler code generation, IR transformations, and kernel scheduling to maximize hardware utilization.

Skills

ML workloads

GPU architecture

C/C++

CUDA/HIP

LLVM/MLIR

Performance profiling

Compiler optimization

Hardware/software co-design

Education

Bachelor’s or Master’s in CS/Engineering

Tools

CUDA

HIP

LLVM

MLIR

Job description

AMD in San Jose, CA seeks a Senior Staff Software Engineer to design and implement compiler optimizations for GPU architectures using LLVM/MLIR. You will develop and optimize GPU kernels with CUDA or HIP, analyze performance bottlenecks, and collaborate across teams to enable new hardware features.

You will build profiling tools and contribute to code generation and kernel scheduling to maximize hardware utilization.