High Performance Computing Software Engineer

Job Responsibilities

• Design and implement high-performance, distributed software solutions for large-scale AI/ML training.
• Optimize low-level system components including Linux kernel, GPU/accelerator kernels, and interconnects.
• Develop and tune communication libraries such as NCCL, MPI, UCX, RCCL, and RDMA-based systems.
• Partner with ML researchers and engineers to support frameworks like PyTorch, MegatronLM, and DeepSpeed in large-scale production environments.
• Contribute to our scheduling, orchestration, and job management systems, including Slurm and Kubernetes.
• Debug and resolve complex issues across the stack—from kernel to container to model.
• Work closely with hardware vendors, upstream open-source communities, and internal teams to drive performance and reliability improvements.

Skills & Experience

• Proven experience developing and optimizing software for large-scale ML workloads (1000+ GPUs preferred).
• Deep understanding of Linux kernel internals and accelerator (GPU) kernel development.
• Proficiency with distributed communication libraries (e.g., NCCL, RCCL, MPI, UCX, SHARP, Libfabric).
• Experience with ML frameworks like PyTorch, TensorFlow, JAX, or MegatronLM.
• Strong knowledge of HPC job scheduling and orchestration tools (e.g., Slurm, Kubernetes, Pyxis).
• Excellent debugging and systems performance tuning skills.
• A collaborative mindset with a focus on shared success and technical excellence.