Senior Software Engineer, AI Triton Kernels

What you'll do

• Design, research, and optimize high-performance matmul, attention, MoE, and fused transformer kernels using Triton.
• Productionize critical Triton/Gluon kernels within vLLM and SGL to ensure correctness and peak throughput.
• Partner with compiler engineers to develop and maintain the Triton AMD backend across ROCm and LLVM.
• Optimize kernel performance across the AMD memory hierarchy, wavefront execution, vectorization, and instruction scheduling.
• Perform rigorous profiling and microbenchmarking using hardware counters to root-cause memory and latency bottlenecks.
• Debug performance and correctness issues across PyTorch, vLLM/SGL runtimes, Triton IR, and the LLVM backend.
• Contribute to the open-source Triton, LLVM, and ROCm ecosystems.

What we're looking for

• Deep expertise in SIMT programming, parallel algorithms, GPU architecture, and performance engineering.
• Ability to perform rigorous quantitative analysis to drive measurable performance improvements.
• Experience designing and optimizing matmul, attention, MoE, and fused transformer kernels using Triton.
• Ability to debug and resolve issues across PyTorch, vLLM/SGL runtimes, Triton IR/MLIR, ROCm, and LLVM AMDGPU backends.
• Deep experience in GPU kernel development, compiler backends, or performance engineering focused on AI/ML workloads (preferred).
• Strong hands-on expertise with Triton, including writing custom kernels and understanding Triton IR lowering (preferred).
• Deep understanding of modern GPU architectures including wavefront execution, memory hierarchy, scheduling, and occupancy (preferred).
• Bachelor’s or Master’s degree in Computer Science, Computer Engineering, Electrical Engineering, or equivalent practical experience (preferred).