Senior Software Engineer, Fabric Networking - GPU

We're seeking a skilled Senior Software Engineer for our GPU Fabric Networking team, building high-performance communication software for demanding workloads in deep learning and HPC.

As a Senior Software Engineer, you will design, develop, and maintain system-level software to support GPU-to-GPU communication. You will collaborate with cross-functional hardware and software teams to build next-generation networking solutions. You will contribute to scalable and reliable GPU fabric architecture for large compute clusters. You will align software development with customer needs and real-world deployment environments.

Responsibilities:

• Design, develop, and maintain system-level software to support GPU-to-GPU communication.
• Collaborate with cross-functional hardware and software teams to build next-generation networking solutions.
• Contribute to scalable and reliable GPU fabric architecture for large compute clusters.
• Align software development with customer needs and real-world deployment environments.

Requirements:

• A degree or equivalent experience in Computer Science, Electrical Engineering, or a related field (B.S., M.S., or Ph.D.).
• 5+ years of professional software development experience.
• Proficiency in C/C++, with strong debugging and system-level problem-solving skills.
• Experience developing software that interacts with hardware and device drivers.
• Solid understanding of system architecture, operating systems, and kernel internals.
• Background in multi-threaded and distributed systems development.
• Experience with Linux development; familiarity with Windows is a plus.
• Strong understanding of networking fundamentals and high-performance interconnects (e.g., InfiniBand, Ethernet).
• Familiarity with OS virtualization technologies like KVM, QEMU, or Hyper-V.
• Comfortable collaborating with a distributed team across different time zones.

Nice to have:

• Experience with the CUDA programming model and NVIDIA GPU architecture.
• Understanding of memory consistency and coherence models.
• Exposure to static/dynamic code analysis, fuzz testing, or fault injection techniques.