About Coaxial AI
We’re driven by the belief that our work can help power the infrastructure behind one of the greatest
economic and industrial shifts humanity will ever see: the AI revolution.
Every AI breakthrough runs on data centers, and the world needs far more of them than it can build today. One of the biggest bottlenecks is the skilled, labor-intensive work of fitting them out. Coaxial AI builds AI-enabled robots that construct data centers, starting with electrical and cooling fitouts, so new compute capacity can come online faster.
The role
You’ll own our robot-learning stack end to end: the policies that make our robots capable, the simulation
and data pipeline that trains them, and the work of getting them running reliably on real hardware.
This is a hands-on individual contributor role. You’ll own outcomes, not headcount. But as one of our
earliest engineers, the technical calls are genuinely yours to make, and your policies will run on live job
sites doing work that shapes how quickly the world can build the infrastructure AI depends on.
It’s a remote role, with occasional travel to job sites to test and deploy.
What you’ll own
• The policies behind our robots’ manipulation and assembly tasks — training, evaluating, and steadily
making them better.
• The simulation and data pipeline that makes policy generation efficient and builds our data flywheel.
• Deployment on real hardware: diagnosing failures on real tasks, closing the sim-to-real gap, and raising
the reliability bar for our autonomy.
• The technical direction of the learning stack, in close partnership with the founding team.
What you’ll bring
• 2+ years developing, fine-tuning, and shipping policies on a real robot or a real task.
• Strong Python and PyTorch (or JAX), with experience training and shipping models.
• Practical experience with modern policy learning, such as imitation learning, reinforcement learning, or vision-language-action (VLA) models.
• Hands-on simulation experience (e.g. Isaac Sim, MuJoCo) to generate and randomize training data,
plus a proven ability to transfer policies to real robots.
• Hands-on experience with real robotic arms, including teleoperation, data collection, and getting
models working reliably on hardware.
• A track record of technical ownership: taking a project from an idea through to something working in the real world.
• Australia based candidates preferred, exceptional candidates located outside Australia also considered.
Nice to have
• Experience with world action models.
• ROS 2 or C++.
• Open-source work or other public results (publications, personal projects) in robot learning.
• Exposure to construction, electrical, or industrial environments.
Compensation
A modest salary now, significant equity from day one, and market-rate pay as we raise. No surprises: we’ll share the numbers and timeline in our first conversation.
