Network Systems Architect
Cerebras builds wafer-scale AI computing systems and a cloud inference platform for training, fine-tuning, and serving AI models.
About Cerebras Systems, Inc.
Cerebras Systems is an AI-infrastructure company founded in 2015. It sells rack-scale wafer-scale computing systems and provides cloud-based, API-accessible AI inference alongside on-premises deployments.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will define multi-generation network architecture and roadmaps. You will derive fabric requirements from workloads, define topology and protocols, evaluate standards-based and custom technologies, use models and prototypes to test designs, and lead specifications and design reviews through implementation and qualification.
Requirements
- 12 or more years of relevant industry experience
- Principal-level technical ownership of a networking, switching, accelerator, or HPC system
- Expertise in low-latency or proprietary interconnects, fabric protocols, or switch architecture
- Knowledge of switch pipelines, buffering, routing, flow or congestion control, and reliability
- Experience deriving network requirements from incomplete workload and system information
- Knowledge of Ethernet, IP, RDMA, BGP, and EVPN
- Experience with implementation, modeling, silicon, lab, bring-up, or debugging work
- Technical decision-making and stakeholder influence
- Knowledge of accelerator interconnects, switch ASICs, NICs, DPUs, transport offload, collective acceleration, coherent memory, or custom fabrics
- Knowledge of NCCL, RCCL, MPI, SHMEM, RoCE, InfiniBand, PCIe, or CXL
- Workload-driven modeling, traffic simulation, emulation, and silicon correlation
- Knowledge of SerDes, packaging, retimers, cabling, optics, and reach constraints
Responsibilities
- Set the architecture and roadmap for scale-up networks and their interfaces
- Derive bandwidth, latency, ordering, availability, and serviceability requirements from workload and communication choices
- Define fabric topology, protocols, switch behavior, routing, buffering, flow control, reliability, and fault containment
- Decide between standards-based technology and custom network designs
- Use performance models, traffic simulation, prototypes, and lab data to test architecture choices
- Write architecture and interface specifications
- Lead design reviews and drive decisions through implementation, bring-up, and qualification