About NerveStax
NerveStax builds AI agents for data teams. They work on the stack a team already runs: dbt, Airflow, the warehouse and git. The agents load data, change models, look after schedules and investigate pipeline alerts.
Every change is tested in a sandbox and reaches a person as a pull request to approve. We are an early, small team in private beta, onboarding data teams now, hosted or self-hosted.
About the role
NerveStax agents change models, operate pipelines and triage alerts on production data stacks. That only works if they are dependable. You'll own the runtime that makes them so: how agents are composed for a job, how tools are described and gated, how state survives a reconnect, and how we measure whether an agent did the right thing. Models will keep changing; the system around them has to keep the guarantees.
What you’ll do
- Own the agent orchestrator and the specialised agents it composes, as the product grows from a handful of jobs to many.
- Design tools with compact outputs, closed choices and risk levels, so the agent has little left to guess.
- Own human approval gates: an agent pauses, a person decides, and the run resumes correctly afterwards.
- Build evals phrased the way people actually ask, scored against what changed in the product, not the transcript.
- Keep adding agents and skills cheap, so a new capability is a definition rather than a rewrite.
- Stay ahead of the model landscape: new models, new providers, and what each one changes about our guarantees.
- Work with the teams using NerveStax to find where agents go wrong, then fix the cause.
We’re looking for people who
- Have shipped LLM-based features to real users and debugged them when they misbehaved.
- Write solid Python and are comfortable with async code, streaming and persistence.
- Measure agent quality with evals rather than a handful of demos.
- Think about failure early: timeouts, retries, partial state and what the user sees.
- Like owning a problem end to end, from design to production.
Nice to have
- Experience with an agent framework such as LangGraph.
- Working knowledge of data tooling: dbt, Airflow, SQL warehouses.
- Built human-in-the-loop or approval workflows.
What we provide
- Competitive salary and meaningful equity.
- Ownership of real problems, from design to production.
- Direct work with the data teams who use the product.
- Your choice of tools, including AI coding agents.
- A say in what the product and the company become.
How to apply
Email us. There is no form, and you don't need a cover letter. Send it to [email protected] with the subject “Founding Engineer, Agents”, and include:
- A few lines about you and why this role.
- Links to work you're proud of: code, writing, talks or a product you shipped.
- Where you are based and when you could start.
We read every email and reply. What happens next ›