Nvidia 6 дней назад
Senior System Software Architect, AI and GPU Networking
- Опыт от 5 лет
- Полная занятость
AIDeep LearningSystem ArchitectureStorageHPCVirtualizationSDNGPU Networking
Описание
Responsibilities
- Enhance NVIDIA
-s GPU Networking offerings for accelerating AI workloads, such as NVIDIA Dynamo, NVIDIA NIXL and NVIDIA UCX, tailored to the unique requirements of AI workloads.
- Co-design hardware features (e.g., in GPUs, DPUs, or interconnects) that accelerate data movement and enable new capabilities for inference and model serving.
- Identify and evaluate new technologies, innovations and partner relationships for alignment with our technology roadmap and business value.
- Lead architecture and design of new technologies and innovations such as runtime systems, communication libraries, AI-specific technologies.
- Lead proof-of-concept development to evaluate and drive such technologies.
Requirements
- Hold a B.Sc., M.Sc. or Ph.D. in Computer Science, Electrical or Computer Engineering from a leading university (or equivalent experience).
- 5+ years of industry experience (or equivalent) in system architecture, AI systems architecture, scaling of AI, parallelism of A
- I frameworks, or deep learning training workloads.
- Experienced in algorithm design, system programming, computer architecture and operating systems.
- Experienced in virtualization, networking and storage.
- Deep understanding of performance profiling and optimization techniques, together with defining and using hardware features.
- Strong programming and software development skills.
- Ability and flexibility to work and communicate effectively in a multi-national, multi-time-zone corporate environment.
Nice to have
- Shown research track record.
- Have experience and passion for system architecture, CPU/GPU/memory/storage/networking.
- Stellar communication skills.
- Knowledge in Deep Learning frameworks and AI communication libraries (NCCL, UCX, MPI and equivalents).
- Deep understanding of Inference and Training workloads and optimizations, like Prefill/Decode, data parallelism, Tensor parallelism, FSDP and others.