Back to jobs
Job Description
- Analyze the networking issues associated with the next generations of GPU hardware, and design, build, and deploy whatever is needed to make them work optimally in our data centers.
- Achieve workload optimal performance NVIDIA Collective Communications Library (NCCL) + Graphics Processing Unit (GPU).
- Compile a comprehensive analysis of performance across different GPU and network generations.
- Determine how customers' ML models will evolve once we have 72 node-NVLink domains.
- Execute full stack optimization for ML networking performance on Google's infrastructure, this spans a wide range, from kernel optimization, user space communication libraries.