Production Engineer, AI Storage for HPC Clusters

MPower Plus; Stati Uniti; Remoto; USD 150.000 - 210.000; Tempo pieno

Descrizione del lavoro

MPower Plus is seeking a hands-on Production Engineer to design, build, and operate storage for AI training across HPC clusters in a fully remote US role. You will connect high‑performance filesystems, object storage, and data management services to AI compute so training data and checkpoints are fast and reliably available.

You will automate provisioning, scaling, upgrades, and recovery with Kubernetes and infrastructure‑as‑code, move data across clusters and clouds, and maintain reliability

Competenze

• Production distributed storage or large-scale data infrastructure experience.
• Strong Linux skills with production debugging abilities.
• Experience writing automation in Python, Go, C++, or Rust.
• End-to-end ownership of reliability or performance improvements.

Mansioni

• Build and operate distributed storage for AI training across clusters and regions.
• Diagnose and optimize storage performance to prevent training stalls.
• Automate provisioning, scaling, upgrades, and recovery using Kubernetes and IaC.
• Migrate data across clusters and clouds without data loss.
• Maintain reliability via observability, testing, safe rollouts, and incident response with cross-team partners.

Conoscenze

Linux

Performance debugging

Python

Go

C++

Rust

Storage systems

Strumenti

Kubernetes

Terraform

Helm

S3 storage

Lustre

Descrizione del lavoro

Role: Production Engineer, AI Storage for HPC Clusters

Location: United States Candidates Only – 100% REMOTE

Job Type: FTE/Permanent

About the role

The customer's Cloud Foundation AI Services team builds and operates the storage that lets AI researchers train models reliably and efficiently across cloud environments. We

connect high-performance filesystems, object storage, and data-management services to AI compute, so training datasets and model checkpoints are fast and available wherever they are needed.

We are looking for a hands-on storage system engineer who can turn demanding AI workloads into reliable, scalable infrastructure. The work spans software development, performance analysis, automation, and production ownership.

What you'll do

• Build and operate distributed file and object storage for AI training: shared researcher filesystems, training datasets, and model checkpoints.
• Diagnose and improve end-to-end storage performance (metadata, throughput, latency, caching, network) so storage I/O stops stalling training.
• Automate the storage lifecycle (provisioning, scaling, upgrades, recovery) using Kubernetes and infrastructure-as-code.
• Move and migrate data across clusters, regions, and clouds without losing integrity or slowing researchers down.
• Keep it reliable through observability, alerting, testing, safe rollouts, and incident response, partnering with research, security, and storage-provider teams.

Must-haves

• Built or operated production distributed storage, high-performance filesystems, or largescale data infrastructure.
• Strong Linux skills, and can debug filesystem, network, and concurrency issues in production.
• Writes maintainable software or automation in a language such as Python, Go, C++, or Rust.
• Has owned a reliability or performance improvement end to end, from spotting the problem through deployment and verification.

Nice to have

• Experience with cloud infrastructure and Kubernetes, or a comparable large-scale orchestration environment.
• Depth in NFS/POSIX and distributed-filesystem internals: metadata, caching, consistency, availability, and recovery.
• Hands-on with one or more distributed or network storage systems such as Hammerspace, Lustre, ZFS/OpenZFS, VAST, or Pure Storage. Depth in comparable systems counts; you do not need every product named here.
• AI or HPC storage patterns: high-throughput data loading, checkpoint writes and restores, metadata-heavy access and staging data for accelerator clusters.
• Infrastructure-as-code and observability tooling (Terraform, Helm, Kubernetes operators; metrics, tracing, benchmarking), and object/S3 storage with tiering and cross-region movement.