Outcome Engineer - Seattle, WA

Architect Multi-Agent Systems: Design the orchestration layers that enable multiple specialized agents to collaborate, debate, and execute complex tasks reliably. You will move us from simple prompts to robust, stateful agentic workflows.

Implement Automated Governance: Create the "guardrails" and policy engines that allow autonomous systems to operate safely at speed. You will build the infrastructure that replaces manual review with programmatic constraints.

Engineer Context and Memory: Build the semantic infrastructure (knowledge graphs, vector stores) that provides agents with deep, structural understanding of our codebase and business intent. You will solve the "blank slate" problem for autonomous workers.

Build Rigorous Evaluation Frameworks: Develop the testing harness that defines success. You will move us beyond "vibes" to deterministic validation of agentic outputs using formal specifications and automated metrics.

Design Self-Healing Loops: Connect observability data directly to agent actions, enabling systems that not only detect failures but autonomously analyze and remediate them.

Prototype and Deploy: We prove success by shipping. You will rapidly prototype new agentic tools and deploy them to production for both our internal teams and customers. This is a creation role; we learn by building.

Experiment and Learn: This is a new field. You will be experimenting with new models, new orchestration techniques, and new ways of validating agent behavior every day.

While we are open to many backgrounds, there are a few baselines for success in this team:

You Must Be a Product Builder: Whether your background is code, design, or leadership, you must have a history of shipping. We don't just theorize; we deploy. You need to understand the visceral difference between a prototype and a product that users rely on.

You Must Be an Active Agentic Builder: This isn't something you read about on Twitter. You built with an agent last week. You're building with one right now. You know the frustration of a prompt that worked yesterday failing today.

You Must Love Intense Collaboration: We are all-remote but high-bandwidth. We leave video links open while we eat lunch. Our pets and kids know our coworkers. If you prefer to go dark for three weeks and emerge with a PR, this is not the team for you. We think, debate, and build together, often synchronously.

You Respect Expertise, But Don't Gatekeep: We are bringing together radically different disciplines. We value deep expertise, but we value the willingness to share it and empower others even more. No silos.

You Must Write to Think: In a distributed, asynchronous, fast-moving environment, clarity of thought is demonstrated through clarity of writing. You share bad ideas to get to good ones. You document your experiments, your architectural decisions, and your failures.

Agentic Stack: Large Language Models, Vector Databases, the usual suspects.

Foundational Stack: Node.js, TypeScript, PostgreSQL, Redis, Kubernetes, AWS.

Notice to Third Party Recruitment Agencies Please note that Onebrief does not accept unsolicited resumes from recruiters or employment agencies. In the absence of an executed Recruitment Services Agreement, there will be no obligation to any referral compensation or recruiter fee. In the event a recruiter or agency submits a resume or candidate without an agreement Onebrief explicitly reserves the right to pursue and hire those candidate(s) without any financial obligation to the recruiter or agency. Any unsolicited resumes, including those submitted to hiring managers, shall be deemed the property of Onebrief.