Principal Deep Learning Communication Architect
$272k - $431k • Remote • US, CA, Santa Clara
Posted 1d ago
About the job
NVIDIA is seeking a Principal Deep Learning Communication Architect to define the long-term technical roadmap for communication libraries across next-generation platforms, ensuring seamless scaling of models to massive clusters. This role involves leading the development of advanced communication primitives and collective algorithms, optimizing for heterogeneous interconnects, and co-designing libraries with application developers to meet the demands of trillion-parameter and Agentic AI. The architect will also collaborate with silicon architects to influence hardware specifications for future networking needs and develop analytical models to predict system behavior under emerging workloads.
Responsibilities
- Define the long-term technical roadmap for communication libraries across NVIDIA’s next-generation platforms.
- Ensure the seamless scaling of models to clusters comprising hundreds of thousands of nodes.
- Lead the development of next-generation communication primitives and collective algorithms.
- Optimize communication for heterogeneous interconnects such as NVLink, Spectrum-X (Ethernet), and Quantum-X (InfiniBand).
- Partner with application developers to architect and implement specialized communication primitives.
- Ensure AI and HPC libraries evolve to meet the requirements of trillion-parameter and Agentic AI.
- Collaborate with silicon architects and software engineers to influence hardware specifications for next-generation networking.
- Develop high-fidelity analytical models and simulators to predict system behavior under emerging workloads.
Requirements
- Ph.D. or M.S. in Computer Science, Electrical Engineering, or a related field (or equivalent experience).
- 12+ years of industry experience in high-performance computing (HPC) or distributed deep learning.
- Deep understanding of 3D parallelism (Data, Tensor, Pipeline) and advanced strategies.
- Deep technical proficiency with NCCL, UCX, UCC, NVSHMEM, or MPI.
- Experience with RDMA, RoCE, and low-level InfiniBand verbs.
- Advanced knowledge of high-throughput inference engines and schedulers, specifically TensorRT-LLM, vLLM, SGLang, and NVIDIA Dynamo.
- Expert knowledge of the NVIDIA GPU memory hierarchy (HBM3e/HBM4, L2 cache) and CUDA programming models.
Benefits
- Competitive salaries
- Generous benefits package
- Equity