Software Engineer, Infrastructure

present of AI; San Francisco (CA); On-site; USD 140,000 - 210,000; Full time

Job summary

Cognition is seeking an experienced infrastructure engineer to run and scale our sandboxed compute, orchestration, and internal tooling for Devin/Windsurf. You will own reliability for large distributed systems, with a focus on safe isolation and robust incident learning.

This role partners with researchers and product teams to ensure the platform meets essential AI workloads. You’ll work on sandboxing, scheduling, and monitoring at scale, gaining close exposure to autonomous coding agents in

Qualifications

• Experience operating large distributed systems.
• Comfortable with Kubernetes, a major cloud and Terraform.
• Writes solid Python.
• Experience with isolation and running untrusted code safely.
• Habit of learning from incidents.

Responsibilities

• Run sandboxed compute, orchestration, internal developer tooling, and monitoring under Devin and Windsurf.
• Operate large distributed systems with reliability.
• Collaborate with researchers and product teams to align system needs with requirements.
• Build and maintain observability, SLOs, and incident response practices.

Skills

Kubernetes at scale

Sandboxing/Isolation

Python systems code

Observability/SLOs

Incident learning

Cloud platforms

Terraform

Tools

Terraform

Job description

The role is about running the systems underneath Devin and Windsurf: sandboxed compute, orchestration, internal developer tooling, and monitoring. A strong candidate has real experience operating large distributed systems, is comfortable with Kubernetes, a major cloud and Terraform, and writes solid Python. Experience with isolation and running untrusted code safely is a clear plus, as is a habit of learning from incidents.

How it fits the AI era

AI agents run code at huge volume, so the infrastructure has to start, isolate and shut down many short and long sessions reliably. Much of the routine setup and debugging can be sped up with AI tools, but judgment about failure modes, capacity and safe isolation still rests with people. Engineers here also act as the partner to researchers and product teams, deciding what the systems truly need.

How it builds AI experience

This job offers close exposure to how autonomous coding agents actually run in production, including sandboxing, scheduling and observing agent workloads at scale. To build toward it, someone could set up a small sandboxed code execution service on Kubernetes, add tracing and alerts, and write up what broke. Using AI coding tools daily on infrastructure work also builds a feel for what agents need from the platform.

Worth showing

• Kubernetes at scale
• Sandboxing and isolation
• Python systems code
• Observability and SLOs

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Cognition

Applied AI lab building Devin, an autonomous software engineering agent that plans, writes, debugs, and deploys code