AI Deep Engineer (Evals & Safety)

CodeSpar builds financial infrastructure for agents that pay and get paid. A person or company defines a mandate, and our systems check the limits before a payment is sent.

We’re hiring an engineer to build evals and adversarial tests for agents working with financial mandates.

You’ll test financial tasks, multi-agent workflows and prompt injection, and check whether agent behavior stays within the mandate.

You should have built and shipped agents, used LangGraph or a similar framework, and be able to turn a failure mode into a reproducible test.

Remote, full-time, Brazil.

Apply through LinkedIn, or email fabiano@codespar.dev with “Evals & Safety” in the subject. Include a short introduction and an example of an agent system, evaluation or security test you’ve built.

https://codespar.dev