About the role
About us Graphcore is one of the world’s leading innovators in Artificial Intelligence compute. It is developing hardware, software and systems infrastructure that will unlock the next generation of AI breakthroughs and power the widespread adoption of AI solutions across every industry. As part of the SoftBank Group, Graphcore is a member of an elite family of companies responsible for some of the world’s most transformative technologies. Together, they share a bold vision: to enable Artificial Super Intelligence and ensure its
benefits are accessible to everyone. Graphcore’s teams are drawn from diverse backgrounds and bring a broad range of skills and perspectives. A melting pot of AI research specialists, silicon designers, software engineers and systems architects, Graphcore enjoys a culture of continuous learning and constant innovation. Job Summary We are seeking for a visionary AI Platform Architect to design and oversee the comprehensive infrastructure stack that powers our most demanding distributed AI workloads. Moving beyond individual hardware components, this role acts as the unifying technical authority across hardware, software, compute, network, and storage. You will be responsible for architecting a cohesive, AI rack scale platform optimized for trillion-parameter LLM training and high-throughput inference. By orchestrating everything from advanced clustering and distributed training frameworks down to the physical layer—spanning PCIe Gen 5/6 pathways, NVMe storage topologies, and RDMA fabrics—you will ensure our AI research and deployment teams have a flawless, frictionless, and extraordinarily powerful platform at their disposal.
Responsibilities and Duties End-to-End Platform Architecture: Define the holistic architecture for highly clustered AI environments, ensuring zero-bottleneck data flow between parallel storage systems, AI compute nodes, and ultra-high-bandwidth network fabrics. Workload Orchestration: Influence the strategy for AI workload scheduling and orchestration, utilizing tools like Kubernetes or Slurm to manage distributed training jobs, model check-pointing, and inference serving at massive scale. Full-Stack Optimization: Profile and eliminate system-level bottlenecks across the entire AI pipeline, tuning everything from deep learning frameworks (PyTorch, DeepSpeed, etc.) down to OS-level NUMA pinning and I/O scheduling. Hardware-Software Co-design: Work closely with software, firmware, and OS engineering to influence platform design, ensuring the software stack fully exploits underlying hardware capabilities, including complex ARM mesh interconnects (RNI, HNF, SNF) and advanced merchant silicon features. Silicon Influencing Strategy: Drive the 3-to-5-year technical vision for the AI platform. Collaborate closely with subject matter experts in processor, memory, storage, GPU, thermal, mechanical, BIOS, and Manageability disciplines to define
