Position 2: Network Solution Architect
Location: San Jose, CA
Duration: 6-12 Months
Job Description:
Client is building a high-performance AI infrastructure platform where network architecture is fundamental to the economics, performance and reliability of the service. Our environment spans data-center backbone connectivity, high-speed cluster fabrics, heterogeneous compute and the systems that move data efficiently between them.
As Network Solutions Architect, you will own the architecture of the networks that connect client's infrastructure. You will translate compute, storage, inference and operational requirements into network designs that are performant, scalable, resilient and manageable in production.
This is a hands-on architecture role. You will make the major technical decisions, define standards and interfaces, evaluate technologies and vendors, and work closely with the engineers who implement and automate the resulting designs.
WHAT YOU'LL OWN
▸ Network Architecture: Define the architecture for client's data-center networks, including backbone, cluster fabric, management and service networks.
▸ AI & RDMA Fabrics: Design topologies for high-throughput AI workloads, distributed inference and large accelerator clusters, including RDMA over RoCEv2 or InfiniBand where appropriate.
▸ Network Standards: Set the requirements and standards for bandwidth, latency, oversubscription, resiliency, routing, addressing, segmentation, QoS and traffic engineering.
▸ Infrastructure Integration: Define the interfaces between the network and compute, storage, inference, Kubernetes and data-center operational systems.
▸ Reference Architectures: Develop repeatable network architectures that deploy across client data-center sites.
▸ Technology & Vendor Evaluation: Evaluate network hardware, software and vendors, and recommend solutions based on performance, reliability, cost and operational complexity.
▸ Architecture to Implementation: Partner with the Network Solutions Developer to turn architecture into production-ready automation, and review designs and implementation plans against client's technical standards.
▸ Measurement-Driven Architecture: Use production measurements to validate decisions, resolve bottlenecks and failure modes, and evolve the network as client scales.
EXPECTATIONS: WHAT YOU'LL BRING
▸ Data-center networking: Deep experience designing production data-center networks: spine/leaf architectures, Ethernet switching, routing, BGP, VLANs/VRFs and segmentation.
▸ High-performance networking: Strong understanding of AI or HPC networking and RDMA, ideally RoCEv2 or InfiniBand, including PFC, ECN and QoS.
▸ AI infrastructure: Understanding of the networking requirements of GPU and accelerator clusters and distributed inference.
▸ Network architecture: A track record of turning application and infrastructure requirements into scalable architectures, standards and implementation plans.
▸ First-principles reasoning about bandwidth, latency, congestion and oversubscription.
▸ Strong understanding of how the network interacts with compute, storage, Kubernetes and distributed applications.
▸ Designing for failure: redundancy, recovery, capacity margins and production operability.
▸ Clear architecture decisions and tradeoffs, communicated to engineers, vendors and leadership.
▸ Design-review discipline, and validating architecture with data rather than assumptions.
NICE TO HAVE
▸ AI data-center fabrics at scale, including production InfiniBand or RoCEv2 deployments.
▸ NVIDIA Spectrum/ConnectX or similar AI networking platforms.
▸ SONiC or other open network operating systems.
▸ Designing networks across multiple data-center sites.
▸ Network automation and programmable networking.
▸ Using agentic coding tools for technical design, analysis or implementation.
Regards
Mark R
Email:
