Product Manager, Networking

Responsibilities

• Own the product roadmap for all internal networking tooling: design automation, provisioning, observability, performance analysis, and incident remediation workflows across frontend, backend, OOB, and BMS networks.
• Drive the strategy and requirements for digital twin tooling that models physical fabric topology, enabling engineers to validate designs, simulate failures, and test config changes before touching production.
• Define and ship BOM generators that produce accurate, version-controlled bills of materials for frontend Ethernet, backend Ethernet, InfiniBand, and OOB networks tied directly to cluster topology specs.
• Own the configuration generation pipeline: translate high-level cluster designs into device-ready configs across switches, routers, and OOB management infrastructure, with correctness guarantees and rollback support.
• Build the observability stack requirements for network telemetry ingestion (gNMI, SNMP, streaming) into dashboards and alerting systems that give operators sub-minute visibility into fabric health and performance degradation.
• Define performance profiling tooling that surfaces InfiniBand and RoCEv2 congestion, all-reduce bottlenecks, and east-west bandwidth saturation at the GPU job level.
• Work with network engineers and site operations to map the full lifecycle of a network event from detection through remediation, then build the tooling that compresses mean time to resolution.
• Partner with infrastructure and software engineering teams to integrate networking tooling into the broader cluster lifecycle: from site design through rack-and-stack, burn-in, and steady-state operations.
• Define the data model and schema standards that sit underneath all networking tools, ensuring BOM data, topology data, telemetry data, and config state are coherent and queryable across systems.
• Conduct working sessions with network engineers, site leads, and operations staff to identify the highest-friction workflows, then prioritize ruthlessly based on operational impact.

Requirements

• 5+ years of product management experience with at least 3 years focused on infrastructure, networking, or platform tooling.
• Direct working knowledge of data center networking technologies: spine-leaf topology, EVPN/VXLAN, BGP, 400G/800G Ethernet, and high-radix switch platforms from vendors such as Arista, Cisco Nexus, or Nvidia Spectrum.
• Hands-on familiarity with high-performance interconnects: InfiniBand (HDR/NDR), RoCEv2, and the operational realities of running large-scale RDMA fabrics under AI training workloads.
• Working knowledge of network telemetry protocols and frameworks: gNMI/gRPC streaming, SNMP, OpenConfig, and at least one observability stack built on top of them (Prometheus, InfluxDB, Grafana, or equivalent).
• Experience shipping internal tooling or developer-facing platform products.
• Ability to write detailed technical specifications that engineering teams can execute against without follow-up clarification.
• Demonstrated track record of reducing operational toil through automation: config generation, provisioning workflows, or similar.

Nice-to-Haves

• Experience at a hyperscaler, neocloud, or large-scale GPU infrastructure company where you owned networking tooling end-to-end (AWS, Azure, GCP, Oracle, CoreWeave, Lambda, or equivalent).
• Prior background as a network engineer or network automation engineer before moving into product: you have personally written Ansible playbooks, Nornir scripts, YANG models, or equivalent configuration automation.
• Familiarity with digital twin or network simulation frameworks: emulated environments built on tools like GNS3, EVE-NG, Containerlab, or proprietary fabric simulation systems.
• Experience defining or operating out-of-band management networks: IPMI/BMC, console servers, and the tooling used to reach devices when the in-band network is down.
• Understanding of BMS integration patterns in hyperscale facilities: BACnet, Modbus, SNMP-based BMS interfaces, and the data normalization challenges that come with multi-vendor BMS environments.
• Exposure to AI/ML workload network requirements: collective communication libraries (NCCL, RCCL), all-reduce topologies, and how fabric decisions impact training throughput and model FLOP utilization.
• Familiarity with network source-of-truth and IPAM systems: NetBox, Nautobot, or internal equivalents used to drive automation.