Requirements

Experienced building and operating a platform used by other engineering teams — you think about your product as a platform, not a one-off tool.
Comfortable making product calls: what capability the platform should own, what it should expose as an API, and when to extend versus build new.
An advanced user of agentic coding tools — you work daily with hooks, MCPs, and context management, even if you haven't built an agentic platform yourself.
Security-minded: you have a working understanding of agent sandboxing, scoping, and redaction, even if it isn't your primary background.
A strong communicator who can document technical tradeoffs clearly for both engineering and non-technical stakeholders.
An advocate for a strong testing culture: familiar with test frameworks, code coverage tooling, CI systems, and post-deployment verification.

What you'll be doing

Design and build validation and safety products used platform-wide: test generation, coverage and safeguards, deterministic and predictive test selection, and mutation testing at scale.
Build the polyglot static analysis layer that gives agent-generated code an automated, language-agnostic first quality gate.
Develop agent-specific safety signals: detect prompt flakiness, evaluate agent trajectories, and score semantic correctness of agent-generated code.
Define clear platform capabilities and make deliberate build-vs-buy, build-vs-extend calls as the platform's scope grows.
Apply sound judgment on agent sandboxing, scoping, and data redaction wherever the platform gives agents access to test infrastructure or execution environments.

You have designed or operated developer-platform infrastructure: distribution, CI/CD, or versioning systems used by many teams.
You have built or integrated LLM-based evaluation systems.
You have worked with mutation testing, static analysis, or test-selection systems at scale.
You have hands-on experience sandboxing or scoping agent access to sensitive systems or data.