Infrastructure Engineer

Orin Labs; San Francisco (CA); Hybrid; USD 180,000 - 240,000; Full time

Job summary

Orin Labs in San Francisco is seeking an experienced engineer to own critical infrastructure for long-running AI agents that operate in the real world. You will design and maintain systems that turning research ideas into deployed models, with a focus on reliability and scale.

You will lead training pipelines, evaluation harnesses, and the runtime/deployment stack, ensuring observability, debugging, and on-call readiness for a complex, long-horizon workflow.

Qualifications

• Experience building and operating production systems that others depend on.
• Strong focus on observability, failure handling, and reliability.
• Ability to own problems from architecture to on-call reality.

Responsibilities

• Own large parts of the training infrastructure that turns research ideas into models.
• Maintain evaluation harnesses that quantify model quality and progress.
• Develop runtime, deployment and observability layers to keep long-running agents performing reliably.

Skills

Production systems

Observability

On-call ownership

End-to-end ownership

Job description

We are making AI that can build megaprojects: power plants, factories, data centers, and the physical infrastructure civilization runs on. Today this work requires large teams coordinating over long timescales. We are making agents that do that coordination directly.

What you'll do

Our agents have been running for years, acting independently in the real world, and that only works if the systems underneath them are rock solid. You'll own large parts of that surface area: the training infrastructure that turns research ideas into models, the evaluation harness that tells us whether they're better, the runtime that keeps long-lived agents executing safely, and the deployment and observability layers that let us see what they're doing and why.

Reliability here isn't a nice-to-have. An agent that drops a commitment, loses state, or silently fails mid-task is worse than no agent at all, so the bar is high and the failure modes are subtle. You'll work hand in hand with research, since most of what we learn only shows up at scale and over long horizons, and shipping looks like infrastructure that makes the whole team faster while quietly getting more dependable over time.

What we're looking for

You've built and operated production systems that other people depend on, and you care about the unglamorous parts: observability, failure handling, and the long tail of things that break at 3am. You're comfortable owning a problem end to end, from the infrastructure design down to the on-call reality of keeping it healthy.

Experience with ML or agent infrastructure (training pipelines, inference, eval harnesses, orchestration of long-running stateful workloads) is a strong plus, but we care more about judgment, ownership, and a track record of making complex systems boringly reliable.

Orin Labs is an equal opportunity employer. We don't discriminate on the basis of race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, genetic information, marital status, military or veteran status, or any other legally protected characteristic.