דרושים»תוכנה» GPU Communication - Team Lead
Description
לפני 5 דקות חברה חסויה Location: Job Type: We are seeking an experienced technical leader to head our collective communication library development team. This role involves leading a team of engineers in developing high-performance collective communication implementations for multi-NPU and multi-node AI workloads. Key Responsibilities Lead the design and development of collective communication primitives (All-Reduce, All-to-All, Gather/Scatter and etc) Architect scalable communication protocols for multi-NPU and multi-node systems Optimize communication performance for NPU architectures Provide technical leadership to the team members in NPU programming, distributed systems, and communication protocols Work with a success-driven worldwide international team (Network, NPU, QA, AI, DL/ML Framework) Define project milestones, deliverables, and technical roadmaps Ensure compatibility with major AI frameworks (PyTorch, TensorFlow, JAX).Requirements: BSc/MSc in computer science/computer engineering or equivalent 8+ years of experience in systems programming and distributed computing 5+ years of leadership experience managing technical teams Expert-level C/C++ programming with focus on performance optimization Experience with NPU programming (Triton / CUDA / HIP / OpenCL) Deep understanding of distributed systems, communication protocols, and network programming Experience with DL/ML frameworks (PyTorch, TensorFlow) and distributed training / inferencing Experience with performance profiling and optimization tools Strong communication and interpersonal skills Preferred Qualifications Experience with NPU communication library development Contributions to open-source projects (PyTorch, TensorFlow, communication libraries) Familiarity with containerization and orchestration Interoperability experience with partners, vendors and external teams.This position is open to all candidates. Hide
Skills
Want AI to find more roles like this?
Upload your CV once. Get matched to relevant assignments automatically.