Staff Platform Engineer - Developer Infrastructure

Our robots are built by a small team that ships fast and deploys onto hardware in customer facilities. Between a laptop and a running humanoid there are: a C++/Python/Rust monorepo, an ARM64 cross-compile matrix, kernel modules and a real-time patch set, container images that have to land on Jetson devices, and a rollout process that cannot brick a robot several timezones away. Today that path is held together by the engineers who also write control code. Your job is to own it so they don't have to. Success is measured in build times, time-to-first-commit for a new engineer, deployment frequency, and the number of infrastructure problems the rest of the team stops thinking about.

This is not a cloud-only role. Roughly half the surface area is physical: lab networks, robot dev boxes, bench and HIL fixtures, on-site LAN infrastructure, and the mirrors that let a deployment site work with no internet at all.

The build graph for a mixed C++/Python/Rust/CUDA monorepo: involving incremental correctness, remote caching, and cross-compilation for AMD64 dev boxes and ARM64 targets.

Our CI capacity: self-hosted runners, GPU and hardware-attached runners, and the queue discipline that keeps PR feedback under ten minutes.

Reproducibility: A build from a tagged commit six months from now should produce a bit-identical artifact.

Versioning, artifact promotion, and the container registry / package mirrors that back it.

Safe, resumable, bandwidth-aware rollout to robots - staged channels, canary robots, and rollback that works over a bad link.

Signing and provenance for anything that lands on a robot.

Reproducible dev environments across laptops, shared dev machines, and robots.

Self-service tooling so an autonomy engineer can get a branch onto a robot without filing a ticket or learning Kubernetes.

Onboarding path: a new engineer builds, tests in sim, and deploys to a bench robot on day one.

Cloud and on-prem compute, storage for multi-TB robot logs, and the training/simulation cluster's operational layer.

Fleet observability: metrics, logs, and traces from robot to dashboard.

Site infrastructure for deployments: VPN/overlay networking, local mirrors, offline-capable registry authorization.

8+ years operating production infrastructure for a software engineering org, with direct ownership of CI/CD or developer platform work.

Deep Linux systems fluency including networking, storage, systemd, kernel and driver debugging.

Hands-on ownership of a major CI system (GitHub Actions, GitLab CI, Buildkite, Jenkins) and an understanding of the build and cache layers underneath it.

Real experience with a C/C++ build system at scale e.g., Bazel, CMake, or equivalent. Should include cross-compilation and dependency pinning.

Infrastructure as code (Ansible, Terraform) and a GitOps mindset: reviewable, reproducible, version-controlled change.

Fluent Python and Bash. You automate rather than document a manual procedure.

Strong written communication. You leave behind decisions that outlast the conversation.