Careers

Build the agentic-native
investigation layer.

Greenfield runtime. Frontier models. No legacy stack, no committees. Ship daily to analysts running real cases. Hybrid in San Mateo.

How we build

Agents are the product.
The whole stack is fair game.

We don't bolt AI onto a workflow tool. We build the runtime, the evals, and the surfaces analysts touch. If a tool is right for the job, we use it. If it doesn't exist yet, we ship it.

Build the agent runtime

Tool interfaces, planners, memory, citations, retries. Design the surfaces investigators will trust.

Frontier-model agnostic

Anthropic, OpenAI, open-weight, in-house. We swap models like libraries, route per task, run our own evals.

Closed-loop learning

Every QA disposition becomes a labeled example. The eval harness and the model fleet are the same product.

Customer access from day one

Watch your code run on a real analyst's screen. Iterate the same week. No PMs between you and the user.

Open roles

5 positions open

View board on Gem

Don't see your role?

We're hiring across engineering, GTM, and operations. If you've shipped investigations or AI infrastructure in regulated environments, we'd like to hear from you.

careers@roe-ai.com

See Roe investigate
your real cases.

Book a demo