OpenAI
Software Engineer, Product Infrastructure - Agent Runtime
- Employment type
- Full-time
- Workplace type
- Hybrid
- Location
- San Francisco
- Published
- Aug 17, 2026
- Last checked
- Oct 26, 2026
About this role
ABOUT THE TEAM
Agent Runtime builds the shared agent primitives and compute runtime that power OpenAI products and applications. As agents take on more complex, longer-running tasks, we build the systems that let them maintain state, use tools, and work reliably in the environments they need.
We work closely with Research, Product, and infrastructure teams to turn advances in model capabilities into useful products. Our work combines deep systems engineering with a clear understanding of customer needs, creating reusable foundations that help teams build and scale new agent experiences.
ABOUT THE ROLE
We’re looking for a software engineer with strong product infrastructure experience to build the services, APIs, and runtime capabilities behind persistent agents. You’ll focus on the product infrastructure that makes agent workflows reliable and scalable, including persistent state, asynchronous execution, orchestration, and visibility into work in progress.
You’ll bring product judgment to infrastructure decisions: understanding how people and developers use agents, translating those needs into clear abstractions, and owning systems from design through launch and production operation. You’ll work closely with partner teams to deliver new capabilities while strengthening the shared foundations they depend on.
IN THIS ROLE, YOU WILL:
- Design, build, and operate backend services, APIs, and reusable runtime primitives that support agent experiences across products.
- Build systems for persistent agent state, asynchronous execution, orchestration, and progress reporting so agents can carry work forward over time.
- Develop infrastructure that connects agents with tools and execution environments, with clear interfaces and safe, reliable behavior.
- Improve reliability, scalability, and performance through observability, production debugging, failure recovery, and thoughtful system design.