Senior Software Engineer, AI Triton Communication

What you'll do

• Design and develop native distributed communication and execution capabilities within the Triton AMDGPU backend.
• Implement Triton compiler and runtime mechanisms for collective operations, remote memory access, and synchronization.
• Optimize performance across compute and communication, including inter-GPU data movement and memory hierarchy utilization.
• Develop distributed Triton kernels and execution models to maximize hardware utilization for AI workloads.
• Analyze, profile, and debug complex cross-stack issues spanning the Triton compiler, ROCm stack, and GPU hardware.
• Co-design next-generation distributed GPU programming capabilities with architecture, compiler, and runtime teams.
• Contribute to the open-source Triton and ROCm distributed ecosystems to drive innovation in GPU computing.

What we're looking for

• Deep expertise in GPU architecture, compiler technologies, and distributed GPU systems is required.
• Proven experience optimizing workloads at multi-GPU scale is required.
• Experience working close to the GPU runtime, communication stack, or compiler backend is required.
• Deep experience in compiler development, GPU software, distributed systems, or performance engineering (preferred).
• Familiarity or hands-on experience with Triton compiler and runtime (preferred).
• Deep understanding of modern GPU architectures, including execution models, memory hierarchies, and hardware performance characteristics (preferred).
• Good understanding of GPU runtime systems, communication stacks, and multi-GPU interconnects like XGMI, NVLink, or InfiniBand (preferred).
• Bachelor’s or Master's Degree in Computer Engineering, Computer Science, Electrical Engineering, or equivalent practical experience (preferred).