GPU Infrastructure Engineer

ornn.com•New York, NY

164d•Onsite

About The Position

Ornn is building the financial infrastructure for AI compute. Our price indices are live on Bloomberg Terminal. We structure and trade compute hedging instruments. And we're now building a platform that brings exchange-grade mechanics like order management, matching, scheduling, and settlement to how compute capacity gets reserved and allocated. We're a lean team in New York, backed by leading venture and strategic investors. You would be one of the first infrastructure engineers responsible for the systems underneath Ornn Fabric. Fabric spans heterogeneous GPU clusters, datacenter infrastructure, schedulers, virtualization, networking, and the control plane that makes physical compute capacity available as a reliable financial product. This is not a traditional cloud infrastructure role. You will work across Linux hosts, Kubernetes and Slurm clusters, containers and VMs, bare-metal provisioning, node management, networking, storage, and distributed control systems. You will work directly with our Head of Engineering and have significant ownership over how Ornn operates infrastructure across multiple datacenters and compute providers. Expect to debug failures across the entire stack, from a process inside a container down to the physical node it is running on.

Requirements

Strong Linux systems knowledge. You are comfortable debugging processes, networking, memory, filesystems, devices, permissions, namespaces, cgroups, and system-level performance issues.
Strong Rust knowledge and experience building production systems software.
Working knowledge of SQL and experience designing or interacting with data-intensive backend systems.
Experience operating containerized and virtualized workloads in production.
Deep familiarity with Kubernetes and/or Slurm and the systems surrounding them. Experience with technologies such as RKE2, Slinky, Docker/containerd, Kata Containers, or similar infrastructure is strongly preferred.
Understanding of virtualization fundamentals including hypervisors, KVM/QEMU-style architectures, virtio, device passthrough, and workload isolation.
Strong understanding of distributed systems, including failure handling, coordination, state management, idempotency, consistency, and observability.
Comfort working across abstraction boundaries. You should be willing to debug an application, kernel interaction, network path, scheduler, hypervisor, or physical server depending on where the problem actually is.
Ability to operate independently in a lean engineering team and own systems from design through production.
BS, MS, or equivalent experience in computer science, computer engineering, electrical engineering, or a related technical field.

Nice To Haves

Experience deploying or operating infrastructure inside datacenters, particularly large GPU or HPC clusters.
Knowledge of modern GPU architecture, including CUDA execution, GPU memory, kernels, PCIe, NVLink, RDMA, GPUDirect, and multi-GPU communication.
Experience debugging GPU workloads and performance at the kernel, runtime, or communication layer.
Understanding of server management and node tenancy architectures, including BMCs, IPMI, Redfish, iDRAC/iLO, out-of-band management networks, and secure tenant access.
Experience with high-performance networking such as InfiniBand, RoCE, RDMA, BGP, ECMP, or high-bandwidth Ethernet fabrics.
Experience with distributed storage systems such as Ceph, WEKA, or other high-performance storage architectures.
Experience building infrastructure for cloud providers, neoclouds, HPC environments, exchanges, or other systems where downtime and incorrect state have direct financial consequences.

Responsibilities

Build and operate the infrastructure layer powering Ornn Fabric across heterogeneous GPU clusters and datacenters.
Design systems for provisioning, scheduling, isolating, monitoring, and managing compute across bare-metal, containerized, and virtualized environments.
Build infrastructure software in Rust and Python for node management, orchestration, telemetry, networking, and cluster operations.
Operate and extend Kubernetes and Slurm environments, including technologies such as RKE2, Slinky, container runtimes, and GPU scheduling.
Design secure multi-tenant compute environments using containers, VMs, hypervisors, Kata Containers, device passthrough, virtio, and related isolation technologies.
Debug complex failures across Linux, networking, storage, schedulers, virtualization, GPUs, and distributed systems.
Build systems for collecting and querying infrastructure state, telemetry, inventory, health, and utilization data.
Automate infrastructure deployment and lifecycle management across clusters rather than relying on one-off operational procedures.
Make reliability a feature: design systems that degrade predictably, recover automatically, and expose enough information to understand failures when they occur.