AI Systems Engineer - AI Platforms - Manager

About This Role

This role is for an AI Systems Engineer to build and operate the foundational substrate that powers EY's AI-native platform. The position involves owning the infrastructure and cloud-native platform layers of the Hybrid AI Multi-Environment Runtime (HAI), from bare-metal and GPU infrastructure through Kubernetes, cluster fabric, and multi-tenant scaling. You will be responsible for building and managing EY Fabric environments across cloud, on-prem, edge, and air-gapped targets, providing the substrate on which EY Agentic AI capabilities run. This is an ideal opportunity for a full-stack infrastructure leader who is comfortable with bare-metal and GPU systems and production Kubernetes at scale, and who prioritizes reliability and portability in regulated client contexts.

Responsibilities

• Own the cluster and cloud-native platform, including compute, Kubernetes and scheduling, cluster fabric/networking, multi-tenancy, and distributed compute, as the substrate for Agentic AI workflows and tooling
• Own the infrastructure foundation, including Ubuntu/OS, BMC/bare-metal, DPU architecture, and NVAIE (GPU/Network/DCGM), ensuring the physical and virtual bedrock is provisioned, patched, and production-ready
• Stand up and manage EY Agentic AI environments across cloud (EKS/AKS/GKE), on-prem AI Factory (RKE2/NVAIE), edge (K3s), and air-gapped deployment modes, maintaining one consistent stack contract across all targets
• Deliver foundational platform capabilities such as Infrastructure Management, Kubernetes & Scheduling, and Cluster Fabric Management, so downstream runtime, data, and execution services can run safely and consistently
• Own cluster lifecycle, autoscaling, GPU pooling/virtualization, and multi-tenancy boundaries (vCluster/Crossplane/Karpenter), providing isolated, elastic capacity per tenant and engagement
• Own secure execution and inference, including Ray Serve, vLLM/NIM/Triton, and NVIDIA Dynamo, with sandboxed execution (gVisor/Firecracker for hosted, NVIDIA Open Shell/vNode for on-prem) for isolated, safe model execution
• Own cognitive and routing, including Envoy AI Gateway, semantic routing (vLLM-SR), model/prompt selection, and streaming response handling, directing each request to the right model under the right constraints
• Collaborate with Dev Ops Engineers on deployment and delivery of the platform itself, including CI/CD/CV (ArgoCD), infrastructure-as-code / Git Ops (Helm/Open Tofu), ensuring environments are reproducible and drift-free
• Own backup, disaster recovery, and cross-environment replication for high availability (Velero, CloudNativePG, Cilium Cluster Mesh), along with patching and platform supply-chain hygiene
• Ensure the substrate is modular and swappable, so components can be replaced without rewriting consumers, minimizing vendor lock-in while preserving the stack contract

Requirements