Principal Engineer, Agent Infrastructure — Scale & Safety

Roblox Corporation; San Mateo (CA); On-site; USD 279,000 - 345,000; Full time

Job summary

Roblox Corporation is seeking a Principal Software Engineer to own Roblox’s agent infrastructure, a core system that runs autonomous AI agents for millions of creators and engineers. You will design, implement, and scale backend systems, ensuring reliability, security, and observability in a production environment.

You will work on agent runtimes, workflows, and sandboxing, collaborating with security and infrastructure teams.

Qualifications

• 8+ years of production backend/infrastructure experience.
• Deep expertise in workflow orchestration or scalable backend infrastructure.
• Hands-on: production code in Go, Python, or Rust.
• Agent-native experience with swarms, loops, and orchestration.
• Experience running agents in production against real systems.
• Bachelor’s degree or equivalent in CS/CE or similar.

Responsibilities

• Own design and implementation of Roblox’s core agent infrastructure.
• Drive reliability and security for long-running agents and sandboxes.
• Raise the bar by evolving agentic workflows and sharing with the team.
• Align with Security, Engineering Efficiency, and Infrastructure so the agent platform advances together.
• Mentor senior engineers, lead design reviews, and push the platform forward.

Skills

8+ years backend infra

Workflow orchestration

Scalable backend

Agentic runtimes

Secure sandboxing

Go

Python

Rust

Agent workflows

Production systems

Kubernetes

AWS

Temporal

Education

Bachelor's degree in CS/CE or equivalent

Tools

Kubernetes

GitOps

Argo CD

Job description

Roblox Corporation is seeking a Principal Software Engineer to own Roblox’s agent infrastructure, a core system that runs autonomous AI agents for millions of creators and engineers. You will design, implement, and scale backend systems, ensuring reliability, security, and observability in a production environment.

You will work on agent runtimes, workflows, and sandboxing, collaborating with security and infrastructure teams.