Principal Site Reliability Engineer

About Sondera

AI agents act on instructions written in natural language, and much of what people mean goes unsaid. A sequence of steps that are each allowed can still do damage. Sondera works on the open problem of semantic integrity: every action an agent takes carries out what an authorized person meant. Enterprise security teams write rules in plain language, and we autoformalize them into policy. Before each action runs, our policy engine renders a verdict given the agent's trajectory: everything the agent has done in the session. The same action, in the same state, under the same policy, gets the same verdict. We publish our research at ICML and NeurIPS workshops and present it at Black Hat Arsenal, BSides and [un]prompted. Our open-source hooks and policy engine already govern coding agents, and enterprise customers run our platform. Our team has built products at Cylance, Databricks, HiddenLayer, NetRise and Obsidian Security, and we're backed by Decibel, Runtime Ventures and Aviso Ventures.

About the role

You own Sondera's production platform: how our services run, ship and recover, and what they cost. Security engineers and CISOs depend on it because their AI agents can't act until Sondera returns a verdict. You decide how a change reaches production and rolls back, what we measure and page on, how we recover from losing a zone or a region, and what an agent may do on its own during an incident. You'll turn our AI security research into enforceable, systems-level controls.

Principal is a senior individual-contributor role; you lead through design and review.

In your first year you will

• Make release and rollback something any engineer on the team can run, for every service.
• Set SLOs for the services that render verdicts, with alerts that page before a customer notices.
• Prove recovery from a lost zone or region with timed drills.
• Build agents that triage incidents from logs, traces and metrics, citing their evidence, and propose fixes a person approves.
• Give every running component an owner, a cost estimate and a budget.
• Keep Sondera deployable in customers' own infrastructure from signed, scanned artifacts.

You may be a good fit if you

• Have run production systems customers depend on, through on-call, and still write code.
• Run containers on Google Cloud, on Cloud Run or Kubernetes, and operate managed Postgres, networking and IAM.
• Write infrastructure as code, and review it as closely as application code.
• Design SLOs and burn-rate alerts from what users experience, and write incident reviews people act on.
• Have led a migration or cutover with a rollback you tested before you needed it.
• Use coding agents daily, and review what they write as closely as you'd review a colleague's work. A person who owns the code approves every change to production.

Prior security experience is useful but not required. Judging when an automated diagnosis can be trusted matters more.

Strong candidates may also have

• Experience debugging Rust or gRPC services in production.
• OpenTelemetry tracing and metrics across services.
• Multi-region Postgres failover and point-in-time recovery.
• Agents or automation for incident triage, with an evaluation of their diagnoses.
• Software that enterprises run in their own infrastructure, shipped as signed Helm charts.
• A public record of your work: papers, talks, open source or technical writing.

Our stack

Rust for the services that render verdicts, Python for AI, and TypeScript for the web applications. Policy compiles to a formal policy language. The platform is defined as code and continuously delivered, and runs in Google Cloud and in customers' own Kubernetes clusters.

How we work

We study four open problems: turning written rules into policy, enforcing that policy at runtime, making agent behavior inspectable, and overseeing fleets of agents, including agents that delegate to other agents, without a person reviewing every action. The people who write our papers also ship the product, so our research meets agents doing real work, and what breaks sets our next question. Whoever builds something helps run it. We build Sondera with coding agents and govern them with it. We settle technical disagreements with evidence. We work remotely: we start from written intent, review each other's changes async, write decisions down where everyone can read them, and flag any decision we can't undo before we make it.

Logistics

• Location: Remote, United States. We meet in person three times a year.
• Compensation: Base salary of $240,000–$280,000 USD, plus equity.
• Benefits: Medical, dental and vision coverage for you and your dependents; a 401(k) plan; unlimited paid time off, and we expect you to take about a week each quarter, plus paid holidays.
• Visa sponsorship: We sponsor visas, including transfers, though not for every role or candidate.
• Interview process: screen → work-sample interview and review → operations interview → career deep dive (a structured walk through your past roles) → references. We aim to make a decision within two weeks of your screen.
• Using AI in your application and interviews: Use AI to prepare, and to polish an application you wrote yourself. In the work sample you'll build without AI, then extend your work with a coding agent, so we see both how you design on your own and how you check an agent's output. The other interviews are without AI.

We encourage you to apply even if you don't meet every qualification.