DevOps Engineer at FieldAI in Boston, MA

Job description

FieldAI is transforming how robots interact with the real world. Our growing R&D team is based in Boston, where we develop risk-aware, reliable, field-ready AI systems that tackle the hardest problems in robotics and unlock the potential of embodied intelligence. We take a pragmatic approach that goes beyond off-the-shelf, purely data-driven methods or transformer-only architectures, combining cutting-edge research with real-world deployment. Our solutions are already deployed globally, and we continuously improve model performance through rapid iteration driven by real field use. About the Role Field AI is transforming how robots interact with the real world. Our R&D team, the FieldAI Research Institute (FAIRI), is based in Cambridge, MA, where we build risk-aware, field-ready AI systems that unlock general purpose intelligence for robotics. FAIRI is looking for a DevOps Engineer to own the infrastructure our humanoid research runs on. Today that infrastructure is borrowed from Field AI's commercial platform teams and held together by researchers doing it in the margins: the humanoid monorepo has no CI, collected robot data lands in S3 and stops there, and there is no registry telling us what any given dataset actually contains. You will be the first dedicated infrastructure hire inside FAIRI. You will build the CI/CD, data pipelines, and IaC that let a small research team ship reliably — partnering with Field AI's platform, cloud, and data-processing teams rather than rebuilding what they already run well. This is a hands-on ownership role, not a coordination role. What You'll Do CI/CD and Build Infrastructure — 30% Stand up CI/CD for the humanoid monorepo: containerize, push to ECR, run unit tests, build, and gate on simulation system tests before promotion. Work with the platform team's self-hosted GitHub Actions runners (ARM, AMD, CUDA, Jetson-class targets) rather than standing up parallel infrastructure. Cut build times through change detection and remote caching — full builds are currently ~45 minutes uncached. Build test infrastructure that lets the same test run against simple sim, Isaac Sim, or real hardware, driven over ROS 2 messages or the robot REST API. Establish per-automation integration tests so shared-library and output-format changes cannot silently break pipelines. Data Pipelines and Orchestration — 30% Stand up and own FAIRI's Airflow stack for humanoid data processing. Build the ingest path from robot to usable dataset: rosbag/MCAP capture, episode segmentation, format conversion, and delivery to training. Implement data lifecycle guardrails — filtering, review-for-deletion, and retention — so idle-robot and failed-run data does not accumulate indefinitely. Build and operate the dataset and mission registry so every dataset is attributable to a subject, session, robot, and purpose. Support MoCapDB in production: ECS Fargate services, AWS Batch retargeting workers, RDS Postgres, and S3, integrated with FieldAI Auth. Cloud, IaC, and Security — 25% Own FAIRI's AWS footprint as code: ECR, S3, IAM roles and cross-account trust policies, VPC and networking, Kubernetes/EKS workloads. Close the gaps where infrastructure is not yet in code, and bring permissions changes under review. Own compliance posture for research tooling — SOC 2 constraints on SaaS, experiment tracking, and data-sharing controls — in partnership with IT and Security. Eliminate person-owned infrastructure: documented owners, runbooks, and access paths for every FAIRI-owned service. Manage secrets, VPN/Tailscale access paths, and hardware-in-the-loop connectivity to robots on the floor. Enablement and Documentation — 15% Write and maintain runbooks, onboarding guides, and architecture documentation so a new engineer can test and deploy on day one rather than learning it from a teammate. Be the interface between FAIRI and Field AI's platform, cloud, and data-processing teams — negotiating what FAIRI reuses versus owns. Support researchers and