Robotics Engineer

About the role

Teach robots to handle fish: wet, slippery and soft. You will own robot learning end to end, until robots work reliably on real hardware, not just in a demo video.

OctaPulse builds computer vision and robotics for aquaculture. Fish farming is one of the fastest-growing food production systems on the planet, but much of its most important work is still done by hand, one fish at a time. Our vision systems already work on fish farms and processing lines. Now we are putting robots next to them.

Fish are a hard thing to handle. They are wet and slippery, they deform when you pick them up, and no two are alike. A rigid gripper and a clean tabletop demo do not get you far. We already have robots working on real hardware. The next job is making them reliable and measured.

You will be one of the first robotics hires. You will own robot learning end to end and work closely with mechanical design. This is not a research role. We care about robots that work on the hundredth trial, not the best video.

The role is on-site in Dallas (Frisco area), with up to 20 to 25% travel to farms and processing sites to see the real work and test where the robots will operate.

What you'll build

- Reliable, gentle handling. Take robot handling from "it worked on a good run" to a measured success rate. Build an honest real-world evaluation, report results the team can trust, understand why things fail, and grow the data that fixes them. Then grow the task toward the messier, faster work real plants need.

- Data collection as infrastructure. Make collecting data, calibrating, and versioning what we learn from boring and repeatable, so every new task starts faster than the last.

- Toward more capable robots. As tasks involve more choices and more variety, help bring newer robot-learning approaches into real work, move from early prototypes to robots built for food plants, and contribute to longer-horizon work on soft, gentle manipulation.

You might be

- Someone who has taught a robot a manipulation task and run it on real hardware, not only in simulation. You can name the task, the data and what broke first.

- Strong in Python, and comfortable reading and patching research code.

- Rigorous about evaluation. You count trials, log failures and do not trust a single good video.

- At home with real hardware: calibration, cabling and the things that come loose.

- Happy to work beside mechanical design. You know the hand often matters more than the software, and you can tell whether a failure belongs to the data, the learning or the hardware.

- Fine with getting wet. You will travel up to 20 to 25% of the time to farms and processing floors, which are cold and loud.

- Able to work on-site in Dallas, TX (Frisco area). This is one of our first robotics hires, reporting to our CTO. It is not a founding-engineer role.

Bonus

- You have worked with recent robot-learning approaches or contributed to open-source robotics tools.