Software Engineer - Agentic SDLC

Who you are

• You have built developer platforms or tooling used daily by a large engineering organisation, with ownership of areas such as:
• build and test infrastructure — build systems such as Bazel or Buck, and CI platforms such as GitHub Actions.
• development environment provisioning — systems such as Coder or GitPod, or built on Kubernetes directly.
• release and deployment systems — CD platforms such as ArgoCD or Spinnaker, and artifact repositories such as Artifactory.
• source control and code review — platforms such as GitHub or GitLab.
• You have shipped agentic systems that do real work inside the software development lifecycle, and are opinionated about what makes them effective and reliable: including but not limited to context design, tool integration, workflow orchestration, and evaluation.
• You treat the platform as a product. You know how to instrument the platform to track both adoption and friction, and you put yourself in front of users regularly to see how things actually work.
• You build with security as a primary constraint. You can reason about compute isolation, identity and access controls, network segregation, and can exercise strong judgement about the trade-off between risk and speed.
• You are excited to work on problems nobody has solved yet. You care about the developer experience, how developers and agents can work together, and you're passionate about building the future of how software gets made.

Who we are

Responsibilities

• Build just-in-time development environments for agents. Sandboxed compute provisioned per task rather than assigned to a person; sized and configured for the work in front of it, ready in seconds and torn down after.
• Make automated code review trustworthy. Review that weighs the whole case for a change — spec, tests, implementation decisions — not just the diff, and decides whether it is safe to ship, whether a human needs to be involved, and what to put in front of them.
• Develop testing frameworks that build confidence in agentic output. Complex, multi-service tests running in realistic environments against real dependencies, proving with evidence what any given change does to system behaviour.
• Establish a chain of custody for every change. An end-to-end trace of where a change came from, what agents acted on it under which credentials, and how it was built — complete enough to verify what it carries and prevent anything insecure from reaching privileged production environments.
• Build evaluation into the platform. Every model upgrade, context change and harness tweak alters how agents behave — measurement allows us to tune the platform on evidence rather than intuition and ensure that the quality of output is improving over time.
• Design build systems that hold up under an order of magnitude more change. Hermetic, incremental builds backed by remote execution and distributed caching — fast enough that agents aren't left waiting, reliable enough that a failure always means something, on capacity that scales with demand.
• Build the context layer agents navigate by. Connecting source, documentation, build and deployment systems and operational signals, so an agent knows which repo owns which part of the system, where a service is deployed and how to build it — enough to take a spec and turn it into a real change on its own.
• Work directly with engineering teams. Understand how they actually build, find where agents stall in their development lifecycle, turn that into platform capability, and guide teams to work more effectively with agents.