Infrastructure Engineer, Sandboxes & Compute

Reinforced Learning; Town of Tonawanda (NY); On-site; USD 120,000 - 180,000; Full time

Job summary

Reinforced Learning is seeking an experienced DevOps/SRE engineer to own the sandbox compute layer, including orchestration, isolation, images and cost. You will manage the sandbox fleet, ensuring fast snapshot, restore and teardown for containerised and VM workloads.

You will enforce isolation between episodes and customers, manage images and dependencies, and optimize cost and performance in cloud-scale infrastructure with IaC and on-call ownership.

Qualifications

• Strong experience with Kubernetes, containers and Linux internals.
• Infrastructure you’ve operated at scale on a major cloud.
• Security-minded: you think about escape paths and blast radius by default.
• Infrastructure as code and on-call ownership.

Responsibilities

• Own the sandbox compute layer: orchestration, isolation, images and cost.
• Manage the sandbox fleet and its scaling.
• Enable fast snapshot, restore and teardown for containerised and VM workloads.
• Ensure isolation between episodes, between customers, and from the outside world.
• Handle images, dependencies and reproducible builds for environment runtimes.

Skills

Kubernetes

Containers

Linux internals

Infrastructure as code

On-call ownership

Job description

The role

Every environment episode needs an isolated, fast, disposable sandbox. We run many at once, for customers who care about isolation. You’ll own the compute layer: orchestration, isolation, images and cost.

What you’ll own

• The sandbox fleet that environments execute in, and its scaling.
• Fast snapshot, restore and teardown for containerised and VM workloads.
• Isolation between episodes, between customers, and from the outside world.
• Images, dependencies and reproducible builds for environment runtimes.
• Cost and performance, watched and improved.

What you bring

• Strong experience with Kubernetes, containers and Linux internals.
• Infrastructure you’ve operated at scale on a major cloud.
• Security-minded: you think about escape paths and blast radius by default.
• Infrastructure as code and on-call ownership.

Helpful, not required

• MicroVMs, gVisor or similar isolation technologies.
• ML or evaluation workloads you’ve supported.