Senior Deep Learning Communication Architect
Core
Architecting communication protocols and optimizing data transfer for distributed deep learning training and inference across hundreds of thousands of nodes.
Role type
Senior Deep Learning Communication Architect
Builds
Scalable distributed training/inference frameworks and high-performance communication systems
Domain
AI/ML Infrastructure, High-Performance Computing (HPC), Distributed Systems
Deliverable
production ML models
Required skills
Distributed deep learning training/inference, Parallelism techniques (Data, Pipeline, Tensor, Expert, FSDP), Communication protocols (MPI, NCCL, UCX, NVSHMEM), C++, Python, CUDA, GPU computing, InfiniBand/RoCE networks
Preferred skills
LLM scaling on large-scale systems, Contributions to DNN training/inference frameworks, Disaggregated serving architectures (Dynamo, Triton)
Technologies
PyTorch, TensorRT-LLM, vLLM, SGLang, NVLink, InfiniBand, CUDA, OpenCL
Responsibilities
Identify and eliminate bottlenecks in data transfer and synchronization; Design efficient communication algorithms and protocols; Collaborate with hardware/software teams on high-speed interconnects; Research and evaluate new communication technologies; Build proofs-of-concept and conduct quantitative modeling.
Seniority
Senior, hands-on IC
