Production Robotics ML Engineer — Deploy ‑ Cells

Role Overview

Relling is building the deployment layer for physical AI. We deploy hybrid robotic cells — classical motion composed with learned manipulation primitives — into specialty manufacturing operations across the defense industrial base. Our cells run today inside our own factory and on partner floors, and the work to make them run reliably under real production conditions is the work we're hiring you to do.

Role Overview

Relling is building the deployment layer for physical AI. We deploy hybrid robotic cells — classical motion composed with learned manipulation primitives — into specialty manufacturing operations across the defense industrial base. Our cells run today inside our own factory and on partner floors, and the work to make them run reliably under real production conditions is the work we're hiring you to do.

In this role, you'll build robotic cells that operate in real production environments with real consequences when they stop. You'll work across the full deployment stack: perception, motion, learned policies, observability, safety, and technician interfaces. The focus is shipping infrastructure that works on a factory floor, not on a benchmark. The work compounds: every primitive you build, every failure mode you instrument, every reliability fix you ship becomes part of the deployment layer every future cell ships on.

What You'll Do

• Build and ship hybrid robotic cells that deploy into specialty manufacturing operations. Cells run on the floor, not in a lab.
• Develop and extend Relling's manipulation primitive library — pick, place, insert, align, trace, press, inspect — that decomposes complex manufacturing tasks into composable, debuggable, reusable sub-skills.
• Productize the integration tooling, safety dossier framework, and observability spine that turns one-off cell bring-ups into configuration against stable interfaces.
• Run cells in continuous production at Fuselage and at partner sites. Diagnose failure modes that only emerge under sustained operation against real customer SKUs.
• Distill frontier vision-language-action models for edge deployment, retaining research-grade behavior on production compute.
• Build the technician-facing diagnostic surface that lets a maintenance technician interrogate a learned-policy cell and resolve drift without external escalation.
• Collaborate with research partners at frontier foundation-model labs to bring their best models into production environments under real operational constraints.

• Experience deploying robotic systems (classical or learned) into production environments with uptime accountability.
• Hands-on experience across the robotics stack: motion planning, controls, perception, state estimation, runtime software, end-of-arm tooling, basic hardware bring-up.
• Strong software engineering and infrastructure skills: data pipelines, training systems, evaluation frameworks, observability tooling, runtime monitors, fault isolation.
• Familiarity with foundation-model robotics (VLAs, imitation learning, RL fine-tuning) and the practical engineering tradeoffs of running them on edge compute.
• Comfort working in safety-critical environments. Familiarity with ISO 10218 / ISO TS 15066 or equivalent industrial safety standards is a plus.
• The ability to move from a failure mode observed on the floor to a fix that ships in days rather than quarters.