Engineering
AI Orchestrator
Remote (US / Canada)Full-time
Substantial embeds with ambitious companies to build intelligence into the way they operate—AI agents, decision systems, and the internal platforms that run on top of them. We stay and run what we build. As an AI Orchestrator, you will be the person who designs how all the pieces fit together. Not a single model or a single prompt—the full system. Which agents handle which decisions, how context flows between them, where humans stay in the loop, and what happens when something fails. This is architecture work, but for intelligence systems that learn, adapt, and operate continuously.
What you will do
- Design multi-agent workflows and orchestration patterns for client engagements—sequencing, branching, fallback, and recovery.
- Choose the right model for the right subtask. A classification step does not need the same capability as a synthesis step. You will own those trade-offs across cost, latency, and quality.
- Build and maintain evaluation frameworks that measure whether orchestrated systems produce good outputs end-to-end, not just at individual steps.
- Implement guardrails between agent steps—content filtering, hallucination detection, confidence thresholds, human-in-the-loop review points.
- Integrate LLMs with external tools, databases, APIs, and enterprise systems. The intelligence has to connect to where the work actually happens.
- Work directly with clients and strategy leads. You will be in the room explaining why the system is designed the way it is, and adjusting when the reality on the ground changes.
What we are looking for
- 3–7 years of experience in software engineering, with at least 1–2 years building production systems that use LLMs or multi-agent patterns.
- Deep familiarity with orchestration frameworks and patterns—LangChain, LangGraph, CrewAI, Anthropic tool use, or equivalent. You have opinions about when to use them and when not to.
- Systems thinking. You can design reliable workflows from components that are individually unreliable. You understand retry logic, context windows, token budgets, and caching at a practical level.
- Experience with RAG patterns, prompt chaining, and context management across multi-step workflows.
- Strong product instinct. You care about whether the orchestrated system actually gets used by the person it was built for—not just whether it passes an eval.
- Clear written and verbal communication. You can draw the system on a whiteboard and explain it to someone who is not an engineer.
Compensation and structure
- Competitive salary, equity, and benefits.
- Remote-first. We have a home base in Seattle but our team is distributed.
- Small team. You will work directly with strategy leads and clients. No layers between you and the problem.
Apply
Interested in this role?
We read every application. If there's a fit, someone from the team will reach out directly—no automated responses.