Principal Software Quality Assurance Engineer

Principal Quality Engineer, Quality Agents and VerificationAbout usWe are a venture-backed startup in stealth, building data infrastructure for the age of AI agents. A small, senior founding team, already working with design partners and early enterprise customers. We value judgment over process and ownership over hand-holding, and we build things that hold up under real load. We will tell you what we are building in conversation, under NDA. The roleYou will own how we know our product is right. This is a principal, hands-on role building the quality layer of an AI-first engineering organisation: the quality agents that test, review and challenge every change, the harness that makes those agents trustworthy, and the shared workflows that put them in every engineer's path from first commit to release. The goal is enterprise-grade quality delivered at the speed agents make possible. We are an AI-first engineering organisation. Engineers orchestrate coding agents for coding, testing, documentation and operations, and stay fully accountable for the result. When code is cheap to generate, verification becomes the constraint. Your job is to make verification fast, rigorous and built into how we work, so that we ship to enterprise customers with evidence rather than hope. You report to our Head of Engineering and will be the first quality engineer on the team, setting the bar for how quality works across the whole stack. What you will do
• Build quality agents. Design agents that generate and extend test suites, review pull requests against our quality bar, run exploratory and adversarial testing, triage failures and flaky tests, and check data correctness in our pipelines.
• Build the harness they run in. Sandboxed environments with scoped credentials, deterministic fixtures and test data, golden datasets, and evaluation suites that measure whether each quality agent finds real defects, how often it raises false alarms, and what it costs per run.
• Design the oracles. Agents write tests cheaply, and they will happily write a test that asserts the bug. You define what correct means for each part of the product, what evidence is sufficient, and what a gate must block.
• Embed quality into shared workflows. Merge and release gates in CI, pre-merge review of agent-generated pull requests, release-candidate checks, and the repository conventions (AGENTS.md, MCP tools) that let every engineer and every coding agent run the same checks locally before they push.
• Own evals for our LLM-facing features. Accuracy, regression, prompt-injection and jailbreak resistance, and latency and cost budgets, measured on every release candidate.
• Keep feedback fast and trustworthy. Suite runtime, flake budgets, escaped-defect rate and change-failure rate, reported openly to the whole team.
• Produce the evidence enterprise customers ask for. Release confidence, regression baselines, and a clear record of what was tested, how, and why it was judged sufficient. Engineers still write their own tests. You own the strategy, the gates and the infrastructure, and you make doing the right thing the easiest path. What you bring
• Ten or more years building software, several of them at staff or principal level owning test strategy, quality infrastructure or developer tooling for production systems with demanding customers.
• A strong software engineer first. You write production-quality code and have built test infrastructure that other engineers depend on every day.
• Deep experience testing distributed, search or data-intensive systems: integration, contract, end-to-end, property-based and performance testing, in places where "correct" is hard to define.
• Hands-on experience building with LLMs: agent design, tool use, evaluation of non-deterministic outputs, and the ways agents fail in real codebases.
• You treat verification as a design problem. Given a system, you can say what the oracle is, what evidence is enough, and what the gate should refuse to let through.
• You already orchestrate coding agents as your default way of working, can show concrete examples, and can tell us what you did not trust them with and how you checked their work.
• Product instinct for internal users. Quality tooling only matters if engineers adopt it. Having retired a tool of your own because nobody used it counts in your favour.
• Systems thinking, high standards, low ego. You disagree hard, then commit fully. Comfort with ambiguity and with the discretion stealth requires. How we work and

what we offer
We measure results and ownership, not hours, and we believe in-person time brings out our best work. Hybrid in Madrid, three office days a week together. Agents do the orchestrated work and humans stay accountable; in this role you design much of the evidence that makes that accountability real. Every iteration puts working software in front of a real user, security is built in, and every non-trivial decision gets an ADR. No assholes, radical transparency on metrics, roadmap and risks, and intensity in seasons rather than permanent burnout. Generous equity: we want early people to share properly in a major success. High-end laptop, private health insurance, 22 working days of holiday. Compensation: a base salary at the top of the Madrid market for principal-level quality and test infrastructure engineers, set by how far your experience goes beyond the bar above, plus an early-employee equity grant well above what is usual for the role. How we hireWe hire for the bar, not the gap. An intro conversation, where under NDA we share what we are building. A hands-on session with a real repo, a coding agent and a stream of agent-generated pull requests, where you design the gate and decide what to check by hand. A deep-dive on a quality system you took from open question to measured result. Founder conversations on customers, trade-offs and values, then references. If this role excites you but you don't match every line, apply anyway. Benefits
• Medical insurance