Principal Software Engineer, Agent Infrastructure

San Mateo, CA · San Mateo · USD 278,530-345,040 / year · 8+ years experience · Full-time · Onsite · Posted Oct 13, 2026

At a glance

The Agent Infra team builds the infrastructure that runs autonomous AI agents for creators and engineers at Roblox.
As an early member of this team, you will design and implement core agent infrastructure to handle long-running, non-deterministic workloads at scale.
You will work on systems that are already in production, solving challenges related to safety, durability, and cost-effectiveness.

Own the design and implementation of core agent infrastructure and scale it for internal and external customers.
Ensure reliability and security for long-running agents through deploys, restarts, and outages.
Run untrusted code in sandboxes while managing credential access and network control.
Define SLOs and observability standards for the platform.
Experiment with agentic workflows and integrate successful approaches into the Foundation AI organization.
Partner with Security, Engineering Efficiency, and Infrastructure teams to align platform development.
Mentor senior engineers and lead design reviews.
Drive the evolution of the agent platform.

8+ years of experience building and operating backend, distributed, or infrastructure systems in production.
Deep expertise in at least one core domain: workflow orchestration, scalable backend infrastructure, agentic runtimes, or secure sandboxing and access.
Proficiency writing production code in Go, Python, Rust, or similar languages.
Experience running agents in production against real systems.
Bachelor’s degree or equivalent experience in Computer Science, Computer Engineering, or a similar technical field.

Security experience with systems that run untrusted code, including sandbox and container isolation (gVisor, Firecracker, microVMs).
Experience with Kubernetes hardening, network segmentation, and egress control.
Experience working with Temporal.
Experience working with AWS.
Experience working with self-hosted models.
Experience building and deploying production systems on Kubernetes using GitOps and Argo CD.
Contributions to open source.
Experience building your own product with real user traction.