Software Engineer – AI/HPC Middleware
Core
Develop and optimize AI/HPC communication middleware (MPI/SHMEM, NCCL/RCCL) over high-performance interconnects for AI and HPC datacenters.
Role type
Senior IC software engineer (HPC/AI middleware)
Builds
High-performance collective communication libraries and middleware stacks for GPU/CPU clusters
Domain
AI & HPC networking, distributed systems, high-performance computing
Deliverable
production ML models
Required skills
C/C++ systems programming, Linux kernel/driver development, MPI/SHMEM, NCCL/RCCL, libfabric/OFI, RDMA concepts, performance profiling and tracing
Preferred skills
Open-source contributions, libfabric providers, Ultra Ethernet (UEC/UET), RoCEv2, congestion control, Omni-Path/OPX
Technologies
MPI, SHMEM, NCCL, RCCL, libfabric, UCX, RDMA, RoCEv2, Omni-Path
Responsibilities
Implement and optimize features for HPC and AI middleware stacks; Integrate middleware with underlying transport layers; Prepare and submit upstream contributions to open-source projects; Collaborate with kernel/driver and switch teams for end-to-end performance validation; Analyze performance traces and reproduce customer issues to deliver fixes
