Senior AI Triton & Distributed GPU Software Engineer

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

Job summary

AMD is advancing Triton on AMD Instinct GPUs, delivering native distributed execution and communication to scale AI workloads across multiple GPUs. The role spans compiler, runtime, and hardware interactions to maximize performance and throughput in data centers and HPC environments.

You will collaborate with architecture, software, and performance teams to push the limits of Triton-based distributed AI on AMD hardware, contributing to open-source ecosystems and ROCm integration while guiding

Qualifications

• Must have hands-on experience in GPU software, compilers, or distributed systems.
• Strong knowledge of modern GPU architectures, memory hierarchy, interconnects.
• Experience with distributed GPU libraries and multi-GPU programming.

Responsibilities

• Design and implement native distributed communication and execution for Triton AMDGPU backend.
• Develop compiler/runtime mechanisms for GPU-initiated communication and distributed execution.
• Improve inter-GPU data movement, overlap of communication and computation, and memory hierarchy usage.
• Optimize distributed Triton kernels and execution models for scalability on AMD Instinct GPUs.
• Analyze, profile, and debug cross-stack issues across Triton, ROCm, and hardware.

Skills

Compiler development

Triton compiler

GPU architecture

GPU runtime

NCCL/RCCL/MPI

Multi-GPU scaling

CUDA/Triton

MLIR/LLVM

Profiling/Debugging

HIP/CUDA/ROCm

AI/HPC workloads

Open-source

Leadership/Communication

Education

Bachelor’s or Master’s Degree in Computer Engineering/CS/EE

Job description

AMD is advancing Triton on AMD Instinct GPUs, delivering native distributed execution and communication to scale AI workloads across multiple GPUs. The role spans compiler, runtime, and hardware interactions to maximize performance and throughput in data centers and HPC environments.

You will collaborate with architecture, software, and performance teams to push the limits of Triton-based distributed AI on AMD hardware, contributing to open-source ecosystems and ROCm integration while guiding