A startup builds software that gives existing mobile robots the task-specific intelligence to run surveys and physically interactive inspections in hazardous industrial facilities. The founders come from robotics research and a decade of deploying autonomous systems in nuclear sites, and they focus on manipulation tasks that are easy to describe (“wipe,” “measure,” “inspect”) but hard to execute reliably: keeping a probe normal to a curved surface, tracing a pipe with a fixed offset, or positioning a detector at precise orientations while respecting manipulator kinematics, joint limits and collisions. They emphasize generating constrained manipulation plans from the geometry the robot observes rather than hand-authoring trajectories, and closing the loop on inspection results so execution success is judged by valid measurements and correct contact, not just reaching a pose.
The technical approach combines AI for semantic interpretation with classical geometry, planning, optimization and joint-level control to provide predictable behavior and safety guarantees in critical environments. The product is hardware-agnostic, mapping procedure specifications onto whatever robot platform a facility has, and is targeted first at radiological inspection in nuclear sites with extensions to oil, gas and hazardous chemical inspections. Go-to-market is direct sales: paid technical validations from tens of thousands to over $100k and deployments up to roughly $500k per robot. Founders invite discussion about where abstraction boundaries should sit between manufacturers, application layers and facility-specific work.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.