BELLEAU LABSAgent Systems Engineer
Atlas does analyst work: reading solicitations, tracing incumbents, building the case for a bid. The agents that do it and the connector that lets a customer run them inside their own AI are one system. You own it.
What you’ll own
- The analyst agents: harness design, tool definitions, and the loops they run.
- The connector that puts our data inside a customer's own AI, and the permissions that keep it honest.
- Evaluation loops. Every change gets measured before it ships.
- Provenance. Every output traces back to a source a customer can check.
- Failure behavior: what an agent does when the record is thin or the question is bad.
What you’ll bring
- You've run agents against real work unattended, and you know the failure modes firsthand.
- Judgment about when a model is the wrong tool.
- Experience with tool-use and retrieval systems past the demo stage.
- Enough security instinct to reason about what an agent should never be able to reach.
Logistics
- Location
- Washington, DC or remote (U.S.)
- Authorization
- U.S. work authorization required
- Engagement
- Full time