Triton Compiler and Kernel Software Engineer

What you'll do

• Develop and optimize Triton compiler support specifically for AMD GPUs.
• Improve compiler lowering, optimization, scheduling, and code generation processes.
• Create high-performance kernels for attention, GEMM, MoE, and other AI workloads.
• Develop and optimize multi-GPU kernels and communication primitives.
• Enable new AMD GPU and interconnect capabilities through effective Triton abstractions.
• Analyze compute, memory, communication, occupancy, register usage, and generated code.
• Resolve complex correctness and performance issues across single- and multi-GPU workloads.
• Contribute designs and implementations to upstream Triton and LLVM/MLIR projects.

What we're looking for

• Bachelor’s or Master's degree in Computer Science, Computer Engineering, Electrical Engineering, or equivalent.
• Experience in GPU architecture and programming.
• Experience in compiler development and optimization.
• Experience in high-performance GPU kernels.
• Experience in multi-GPU communication and collective operations.
• Experience in distributed AI training and inference.
• Experience in AI workload and framework performance.
• Experience in low-level performance analysis.
• Experience with AMD GPU architecture and ROCm (preferred).
• Experience optimizing kernels with Triton, HIP, CUDA, or GPU assembly (preferred).
• Experience with Triton, LLVM, MLIR, or another optimizing compiler (preferred).
• Knowledge of GPU execution models, memory hierarchies, synchronization, and instruction pipelines (preferred).
• Experience with collective communication, distributed programming, and libraries such as RCCL or NCCL (preferred).
• Understanding of communication topologies, interconnects, synchronization, and communication–computation overlap (preferred).
• Familiarity with distributed training, inference, tensor parallelism, expert parallelism, or pipeline parallelism (preferred).
• Experience using AI coding tools and autonomous agents to accelerate software development (preferred).
• Familiarity with AI primitives, reduced-precision formats, and performance profiling (preferred).
• Contributions to complex or open-source software projects (preferred).
• Strong analytical, debugging, communication, and collaboration skills (preferred).