OJOx is building the data layer for construction robotics.
We capture what should be built together with how skilled humans build it, so the same interface can guide workers today and train robots tomorrow.
See how OJOx connects design intent, human perception, physical action and robot-compatible data in one captured experience.
Construction robotics isn’t just a more-data problem. It’s a right-data problem. Most physical-AI demonstrations capture what people see and what they do, but not the intended physical state that made those actions correct.
OJOx captures what should be built together with how skilled humans make it real. The same interface can guide people through complex work today, capture how that work is performed, and create the demonstrations needed to teach machines tomorrow.
Guide the human. Capture the skill. Teach the machine.
Read our missionMost demonstration data captures observation and action. OJOx also captures the intended state those actions are trying to realise.
Design intent, egocentric perception, spatial context, whole-body motion and hand motion remain connected in one synchronized record.
OJOx places the intended state directly into the worker's view, aligned to the physical workspace.
The same interface can help people understand unfamiliar assemblies and complex tasks while capturing how the work is actually performed.
Humans do not disappear from the loop on day one.
OJOx connects skilled human motion to humanoid embodiments through teleoperation and retargeting, creating a path from guided human work to robot-compatible demonstrations.
A single OJOx session can record what the worker sees, what should be built, how the body moves, how the task evolves and how that motion transfers to a humanoid representation, all on one timeline.
OJOx sits at the intersection of construction robotics, human demonstration data and embodied intelligence. Our research asks how design intent can become part of the data robots learn from.
A demonstration record that carries the specification the worker was building toward, registered to the workspace, visible in their own view and synchronized with how they moved.
Read the paperMohamed Dawod, Sean Hanna · Construction Robotics. The peer-reviewed foundation this work builds on: using the building model as a prior for robotic perception on site.
View publicationWe’re opening early access to teams working in construction, industrial assembly, humanoid robotics and physical AI.
Whether you want to guide workers through complex tasks, capture skilled physical work, or explore training data for humanoids, we’d like to hear from you.