Staff Networking Platform Engineer, Infrastructure

Groq•San Francisco, NY

About The Position

Groq is seeking a networking engineer to own networking across GroqMetal’s management plane and tenant clusters. The role encompasses network architecture, configuration, security, and production operations, from host and Kubernetes networking through data center fabrics, edge peering, and interconnects. The engineer will manage network changes through fabric managers, automation, and GitOps, applying detailed knowledge of protocols, packet paths, and failure behavior to deliver reliable customer connectivity, isolation, and performance as managed platform capabilities. The role includes building Kubernetes-native automation for tenant and node network lifecycles; managing routing, addressing, service exposure, and GPU networking; and maintaining consistency among intended configuration, applied state, and network inventory. The engineer will establish standards and declarative configuration practices, verify tenant isolation through lifecycle transitions, and own telemetry, troubleshooting, and incident remediation in production. Success means customers can provision and use networking reliably, tenant boundaries are demonstrably enforced, and network changes and failures can be diagnosed without manually reconstructing the system.

Requirements

Deep networking knowledge and production ownership, including designing and operating data center networking services through deployment, expansion, maintenance, and incidents.
Strong understanding of TCP/IP, BGP, routing policy, peering, ECMP, segmentation, and interconnects, with the ability to explain packet paths, diagnose failures, and assess configuration-change consequences.
Bare-metal data center networking expertise, including leaf-spine fabrics, underlay and overlay networks, and tenant separation using technologies such as VLANs, VRFs, and EVPN/VXLAN.
Strong Kubernetes and Linux networking experience, including practical knowledge of CNI behavior, Cilium or comparable systems, network policy, service exposure, and the interaction between cluster networking and physical infrastructure.
Kubernetes-native development, network automation, and GitOps experience; experience building Kubernetes controllers, operators, and custom resources in Go.
Ability to integrate network APIs through reconciliation loops that safely handle retries, partial failures, and resource cleanup.
Experience managing changes through version control, code review, automated testing, and controlled rollout.
Experience building multitenant infrastructure or managed Kubernetes products, including customer onboarding, provisioning, upgrades, scaling, and production support.
Ability to turn customer requirements into supported features with defined behavior, limits, and operational procedures.
Experience securing networks that run customer workloads, including designing and operating tenant separation, management network boundaries, and traffic controls, with practical methods for testing those controls.
Fabric management and IPAM experience with Netris, UFM, NetBox, or comparable systems.
Product judgment grounded in networking: ability to understand customer requirements, evaluate network implications, and define capabilities the platform can reliably deliver and support.
Must meet applicable export-control eligibility criteria; for the U.S. posting, candidates must qualify as U.S. Persons for export-control purposes or otherwise be eligible for an applicable export license.

Nice To Haves

NVIDIA GPU infrastructure and AI networking expertise, including InfiniBand, Spectrum-X, NVL72 systems, RDMA, RoCE, and congestion control.
Networking experience for large GPU clusters, including distributed workload performance, high-bandwidth east-west traffic, and multi-site connectivity.

Responsibilities

Translate customer connectivity, isolation, access, and performance requirements into supported platform features, and define how those features are requested, provisioned, changed, observed, and removed.
Build Kubernetes-native automation to create, modify, and remove tenant networks and manage node membership through provisioning, scaling, repair, replacement, and reassignment.
Verify isolation and connectivity at lifecycle transitions, recover safely from partial failures, and remove stale configuration before resources are reassigned.
Own dynamic configuration of tenant networks, BGP, routing policy, VLANs, and overlays across management, tenant, and external networks.
Use Kubernetes-native automation and fabric-manager APIs to provision and manage customer service exposure, ingress and egress, VIPs, load balancers, and public IPs; validate convergence, failover, reachability, and isolation.
Own networking for the Groq management plane and tenant clusters, including Cilium, network policy, service routing, ingress, egress, and integration with bare-metal networks.
Automate tenant configuration and lifecycle operations for GPU networking, including NVL72 systems, InfiniBand, RoCE, and GPUDirect, through supported platform and fabric-management interfaces; validate connectivity, isolation, and performance as allocations and node membership change.
Own dynamic address allocation, network inventory integration, and tenant network configuration through IPAM, NetBox, and fabric-manager APIs such as Netris and UFM; keep intended configuration, applied state, and inventory consistent.
Own segmentation and traffic policy across the management plane, tenant clusters, and physical fabrics; define permitted communication and verify isolation through provisioning, scaling, maintenance, and hardware reassignment.
Work with security engineering on network access controls and hardening requirements.
Establish addressing, routing, and configuration standards for the management plane and tenant networks.
Implement declarative configuration and continuous reconciliation through Kubernetes-native controllers, fabric-manager APIs, and GitOps, with code review, automated testing and validation, staged rollout, and recovery procedures.
Detect and reconcile configuration drift, maintaining tenant isolation as nodes enter, leave, or return to tenant networks during provisioning, repair, and reassignment.
Own network telemetry, troubleshooting, and incident remediation; diagnose packet loss, routing failures, congestion, and unexpected traffic across hosts, clusters, fabrics, and the edge.
Turn recurring incidents and performance findings into prioritized improvements to the managed service.