Software Engineer, Infrastructure

Cognition Cognition · Coding AI · San Francisco, CA · Research & Development

Software Engineer, Infrastructure role at an applied AI lab building end-to-end software agents (like Devin, the AI software engineer). The role focuses on building and operating the core infrastructure for agent execution, including compute, orchestration, networking, and platform systems. This involves owning agent execution infrastructure (sandboxed compute environments, VM orchestration, container management), building the developer platform (CI/CD, deployment systems), driving reliability and observability, and scaling infrastructure with product growth. Requires deep systems engineering, cloud/container expertise (Kubernetes, AWS/GCP/Azure), strong Python skills, observability instincts, and a security/isolation mindset for agentic workloads.

What you'd actually do

  1. Own agent execution infrastructure: Design and operate the sandboxed compute environments that power Devin's task execution, including VM orchestration, container management, resource scheduling, and isolation at scale.
  2. Build and maintain the developer platform: Create the internal infrastructure that Cognition engineers depend on every day: CI/CD pipelines, deployment systems, developer tooling, and the abstractions that let product teams ship fast.
  3. Drive reliability and observability: Define SLOs, build monitoring and alerting systems, lead incident response, and close the loop on postmortems so the same failure never happens twice.
  4. Scale with the product: Anticipate capacity and architecture needs before they become bottlenecks; make deliberate infrastructure investments that stay ahead of product growth.
  5. Partner across engineering: Work closely with product engineers and researchers to understand what the systems they are building require, and design infrastructure that meets those needs without creating unnecessary complexity.

Skills

Required

  • Deep systems engineering
  • Cloud and container expertise (Kubernetes, AWS, GCP, or Azure, Terraform)
  • Strong software engineering fundamentals (Python)
  • Observability instincts
  • Security and isolation mindset
  • Experience with sandboxing, network isolation, and secure multi-tenant compute

Nice to have

  • Relevant industry experience (frontier AI lab, applied AI company, developer tools company)

What the JD emphasized

  • agent execution infrastructure
  • sandboxed compute environments
  • agentic workloads
  • security and isolation mindset

Other signals

  • building end-to-end software agents
  • AI that can reason on real-world tasks
  • agent execution infrastructure
  • developer platform for AI engineers
  • security and isolation for agentic workloads