Onyx Automation mark
Onyx AutomationRobotics and AI Automation StaffingAn Onyx Staffing Company
 Hire talent

Insights

Notes from inside robotics hiring.

Short pieces for founders, hiring managers, and TA partners who are tired of generic “AI talent” advice — and tired of agency sites that lead with a 25,000-person database.

Why robotics hiring still breaks at Staff-plus

Most robotics orgs can hire a strong mid-level engineer who has shipped on ROS. The stall happens when the spec quietly becomes “owns the stack, mentors the team, talks to customers, and still writes the hard node.” That is three jobs. Candidates who can do it are not lurking in easy InMail lists — they are inside a small set of companies, often unresponsive to generic outreach, and they will test whether you understand the difference between a demo and a fleet. The fix is almost never “more pipeline.” It is a cleaner scope, a calibrated peer on the loop, and a recruiter who will not waste a Principal’s Thursday on a mid-level interview plan.

Autonomy vs. applied AI: two different candidate pools

LinkedIn will happily show you people tagged with both “autonomy” and “LLM.” Treat that overlap as a warning, not a feature. Classic autonomy talent grew up on estimation, planning, safety cases, and ugly integration. Applied AI talent grew up on data, evals, and models that may never have touched an actuator. Both are valuable. Mixing the req — “we need someone who can train a VLA and also own ISO 26262” — is how a search burns a quarter. Pick the bottleneck. Hire for it. Rent or partner for the rest.

Interview loops that do not scare off hardware-aware ML talent

A six-hour coding gauntlet designed for a backend platform hire will filter out the scientist who has actually put a policy on a robot. Useful signals: a short design conversation about latency and failure modes, a look at how they think about data from the fleet, and time with the people they would sit next to. Take-homes should be bounded and paid if they ask for a weekend. Onsite robot time beats another leetcode round. The goal is not to be “nicer.” It is to measure the work you will actually need done.

Physical AI is a hiring problem, not a branding problem

“Physical AI” and “embodied AI” are how the market now describes models that have to survive contact with the world. The hiring implication is simple: you are competing for people who have lived in both a training loop and a hardware constraint. Firms that staff “AI” and “robotics” from the same generalist desk will keep sending you adjacent talent. Adjacent is how quarters disappear. Spec the bottleneck. Hire for it. Rent or partner for the rest.

Chatbots, sourcing, and what still requires a human

Tools got better. The part that did not change: senior robotics people take meetings with recruiters they trust and ignore the rest. Automation can clean a list. It cannot explain why this particular humanoid company is not a circus, or why a medtech robotics role is worth leaving a lab. That is still a conversation. We use new-school tools. We do not outsource judgment to them.

Want this conversation about your req?

Bring the job description. We will mark it up.

Book a call