About the job
About Sygaldry Technologies
Sygaldry Technologies is at the forefront of innovation, creating quantum-accelerated AI servers designed to dramatically enhance the speed of AI training and inference. By merging quantum computing with artificial intelligence, we aim to revolutionize the landscape of superintelligence while tackling the challenges posed by escalating compute costs and energy constraints. Our AI servers uniquely integrate various qubit types within a robust, fault-tolerant architecture, providing the essential elements of cost-effectiveness, scalability, and speed required for cutting-edge AI applications. We are pioneers in the realms of physics, engineering, and AI, confronting the most formidable challenges with a culture grounded in optimism, rigor, and collaboration. Join us in defining the intersection of quantum and AI to drive significant global advancements.
About the Role
As a Senior Systems Architect, you will play a pivotal role in shaping the architecture of our server systems, facilitating connections to GPU/XPU scale-up networks, and enabling quantum computing subsystems to amplify production training and inference workloads exponentially. This role requires you to design, analyze, and model the interconnections, synchronization, and scalability of diverse compute elements, with a primary focus on GPU scale-up fabrics, collective communication, and identifying system-level performance bottlenecks. You will engage in hands-on architecture design, performance modeling, and lightweight simulation/emulation, influencing hardware, firmware, and software co-design decisions directly.
Your Responsibilities
- Architect comprehensive end-to-end heterogeneous compute systems integrating CPUs, GPUs, accelerators, and quantum-adjacent subsystems.
- Analyze scale-up interconnects (e.g., NVLink, UALink, Infinity Fabric) and evaluate their effects on latency, bandwidth, and synchronization.
- Design timing arcs, critical paths, and throughput models for complex multi-device systems.
- Identify system-level bottlenecks in training, inference, and reinforcement learning workflows at hyperscaler scale.
- Model GPU collective operations (e.g., NCCL, RCCL) across nodes, racks, and pods.
- Evaluate interaction patterns of collective communication with emerging accelerators and non-GPU compute elements.
- Collaborate closely with software and firmware teams to ensure architectural assumptions align with practical implementations.
Cross-Functional Collaboration
Act as an essential link between various teams, fostering collaboration and innovation across the organization.
