Senior Deep Learning Communication Architect
$184k - $357k • Remote • US, CA, Santa Clara
Posted 1d ago
About the job
NVIDIA is seeking a Deep Learning Communication Architect to join its software architecture group. This role involves scaling Deep Neural Network (DNN) models and training/inference frameworks to systems with hundreds of thousands of nodes. You will be instrumental in optimizing communication performance by identifying and eliminating bottlenecks in data transfer and synchronization. The position requires designing efficient communication protocols tailored for deep learning workloads and collaborating with hardware and software teams to craft systems that leverage high-speed interconnects and communication libraries. You will also research and evaluate new communication technologies to enhance performance and scalability, and develop proofs-of-concept to validate new strategies.
Responsibilities
- Optimize communication performance in distributed deep learning training and inference.
- Design and implement efficient communication algorithms and protocols for deep learning workloads.
- Collaborate with hardware and software teams on system design for high-speed interconnects and communication libraries.
- Research and evaluate new communication technologies for deep learning systems.
- Develop and implement solutions, including proofs-of-concept and quantitative modeling.
Requirements
- Ph.D., Masters, or BS in Computer Science, Electrical Engineering, or a related field, or equivalent experience.
- 6+ years of experience in building, scaling DNNs, or deep learning training/inference workloads.
- Experience evaluating, analyzing, and optimizing LLM training and inference performance on cutting-edge hardware.
- Deep understanding of parallelism techniques (Data Parallelism, Pipeline Parallelism, Tensor Parallelism, Expert Parallelism, FSDP).
- Understanding of emerging serving architectures (Disaggregated Serving, Triton).
- Proficiency in developing code for DNN training/inference frameworks (PyTorch, TensorRT-LLM, vLLM, SGLang).
- Strong programming skills in C++ and Python.
- Familiarity with GPU computing (CUDA, OpenCL) and InfiniBand/RoCE networks.
Benefits
- Equity
- Benefits package