Sr. Software Development Engineer
Core
Research, design, develop, and test high-performance collective communication libraries, GPU communication runtimes, and benchmarking tools for AI/HPC workloads on AMD Instinct GPUs.
Role type
Senior IC software development engineer (GPU/HPC systems)
Builds
Open-source libraries for GPU communication (RCCL) and benchmarking tools (TransferBench) for distributed training and inference frameworks.
Domain
Semiconductor / High-Performance Computing (HPC) / AI Infrastructure
Deliverable
production ML models
Required skills
Python, C, C++, CUDA, GPU software programming, benchmark development, software debugging (Linux/Unix/macOS), data structures, multi-threading, collective algorithms (AllReduce, AllGather, etc.), RDMA/RoCE/InfiniBand protocols, PyTorch, JAX, DeepSpeed, vLLM
Preferred skills
Experience with graduate-level coursework, research, or internship in GPU architecture or parallel computing
Technologies
ROCm, RCCL, TransferBench, PyTorch, JAX, Megatron-LM, DeepSpeed, vLLM, CUDA, RDMA, RoCE, InfiniBand, PCIe, Infinity Fabric
Responsibilities
Design and optimize multi-GPU and multi-node communication across ROCm stack; Develop and tune collective algorithms for distributed training/inference frameworks; Design micro-benchmarks to evaluate GPU-to-GPU and host-to-device transfer performance; Collaborate with hardware, firmware, and AI framework teams to validate interconnect performance and resolve cross-stack issues.
Seniority
Senior, hands-on IC
