דרושים»תוכנה» GPU Senior Software Engineer
Description
לפני 5 דקות חברה חסויה Location: Job Type: We are seeking a skilled software engineer to join our NPU software stack development team. This role involves developing high-performance GPU programming frameworks, runtime systems, and libraries for AI/ML workloads. You will be responsible for implementing, optimizing, and maintaining GPU software stack components to support distributed AI training and inference. Key Responsibilities Identify bottlenecks, analysis and optimize in distributed NPU eco-system Design and develop NPU memory management system Design and develop optimized NPU development framework, execution path and debugging Develop compatibility with AI frameworks (Triton, PyTorch, JAX) Write high-quality, well-tested code with comprehensive documentation Collaborate with other teams (Hardware, Network, QA, AI Framework Integration) Participate in code reviews and technical design discussions.Requirements: 5+ years of experience in distributed system programming 3+ years of experience with NPU programming (Triton, CUDA, HIP, OpenCL) Expert-level C/C++ programming with focus on performance optimization Expert-level Python programming with focus on DL/ML frameworks (PyTorch/JAX/etc) Deep understanding of NPU architecture, memory tiering, and programming models Knowledge of NPU runtime systems Experience with performance profiling and optimization tools Strong problem-solving and debugging skills Experience with version control systems, Ticking system and collaborative development Team player with excellent communication skills Fast learner, highly organized, detail-oriented with high motivation Preferred Qualifications Experience with NPU software stack development Experience with large-scale NPU systems (100+ NPUs) Experience with DL/ML workloads (oriented AI) and distributed training / inferencing Familiarity with containerization and orchestration.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.