A leading AI technology company is seeking a foundational engineer to lead GPU Networking efforts. This role involves architecting software for unifying GPUs, integrating high-performance networking capabilities, and optimizing large-scale AI models. The ideal candidate will possess deep experience with InfiniBand and RoCE, fluency in C++ or Python, and the ability to enhance system performance. Competitive compensation and benefits are offered, including equity and comprehensive insurance coverage.
Qualifications
Deep experience with InfiniBand, RoCE v2 protocols.
Fluency in C++ or Python, with knowledge of NVIDIA hardware.
Experience with networking performance characterization.
Ability to optimize data movement and system performance.
Responsibilities
Integrate RDMA/RoCE capabilities into our inference stack.
Tune networking layers for efficient communication.
Enable fast startup speeds for large models.
Design tools for visualizing packet flow and performance.
Develop custom communication kernels.
Skills
High-performance networking protocols
C++ or Python
Memory hierarchy in NVIDIA architectures
Custom C++/Python bindings
Debugging NVLink topology
Tools
NCCL
NVSHMEM
UCX
GPUDirect Storage
Job description
A leading AI technology company is seeking a foundational engineer to lead GPU Networking efforts. This role involves architecting software for unifying GPUs, integrating high-performance networking capabilities, and optimizing large-scale AI models. The ideal candidate will possess deep experience with InfiniBand and RoCE, fluency in C++ or Python, and the ability to enhance system performance. Competitive compensation and benefits are offered, including equity and comprehensive insurance coverage.