Senior Software Developer - Network and Collectives
Core
Design and implement collective algorithms and topology-aware transport layers for an AI accelerator's scale-out communication substrate.
Role type
Senior IC systems software engineer (HPC/collective communication)
Builds
High-performance collective communication libraries (all-reduce, all-gather, etc.) for multi-node AI training/inference
Domain
AI hardware systems, High-Performance Computing (HPC), Distributed Systems
Deliverable
production ML models | infrastructure
Required skills
C++, concurrency, lock-free design, collective libraries (NCCL/MPI), performance profiling, bandwidth/latency tuning
Preferred skills
topology algorithms, congestion control, RDMA/InfiniBand/RoCE, GPUDirect, large-scale distributed training/inference stacks (vLLM, SGLang, TensorRT-LLM)
Technologies
C++, NCCL, MPI, InfiniBand, RoCE, RDMA
Responsibilities
Design collective algorithms tuned to interconnect topology; Build topology-aware transport layers; Optimize end-to-end collective performance; Co-design with HW and Multi-Node Runtime teams; Ensure correctness and numerical determinism of reductions
Seniority
Senior, hands-on IC
