Staff Runtime Engineer

About the role

What you'll do

• Own the design and evolution of the agent task execution engine, including scheduling, retry logic, cancellation, and concurrency controls.
• Define and enforce SLOs for task throughput, latency, and error rates across production deployments — and own the on-call rotation for runtime incidents.
• Build and maintain the observability layer: structured telemetry, distributed tracing, and the dashboards that let both our engineers and customers see what their agents are doing in real time.
• Lead architectural reviews for new runtime capabilities, ensuring changes are backward-compatible, well-instrumented, and tested against adversarial workloads before they reach production.
• Partner with the security and AppSec teams to harden the runtime against prompt injection, runaway tasks, and resource abuse vectors that are unique to agentic systems.
• Mentor platform engineers, review critical code paths, and raise the team's standard for system design — particularly around distributed state and failure modes.

What we're looking for

• 8+ years of backend engineering experience with at least 3 years in a senior or staff-level role owning distributed systems in production.
• Deep familiarity with task queue patterns, orchestration engines (Temporal, Celery, BullMQ, or similar), and the failure modes that come with them at scale.
• Strong understanding of observability primitives: structured logging, metrics, distributed tracing, and how to build dashboards that surface signal over noise.
• Experience defining SLOs, running incident retrospectives, and making architectural changes based on production data rather than intuition.
• Comfortable writing TypeScript or Go (our primary runtime languages) and reviewing code across language boundaries.

Nice to have

• Prior experience building infrastructure for AI workloads — especially LLM inference coordination, agent orchestration, or multi-step pipeline execution.
• Familiarity with Vercel Edge, Cloudflare Workers, or other edge compute environments where cold starts and memory constraints are real constraints.
• You have shipped a runtime or platform component that is now used by customers you have never met and still works correctly.

Interview process

1. 1

   Intro call (30 min)

   A conversation with our engineering lead to understand your background and what you're optimizing for in your next role. No coding, no tricks.

2. 2

   System design interview (60 min)

   We'll give you a real problem from our runtime and work through a design together. We care about your reasoning, not whether you arrive at our exact answer.

3. 3

   Technical deep-dive (60 min)

   Walk us through a system you're proud of. We'll ask hard questions about the tradeoffs you made and what you'd do differently today.

4. 4

   Offer and references

   If we're aligned, we'll move quickly. References are a conversation, not a checkbox.