Robots that learn from your best people.
Autonomous skills for extreme environments.
Evoprax records how skilled workers use their hands and where they look, with XR headsets, and turns it into training data for humanoid robots in heavy and hazardous industry.
The problem
Humanoids are arriving. Skills are not.
Robot makers can build the body. What they lack is data on how real experts do real manual work: the grip, the order of steps, the glance before a risky move. On every site that knowledge sits with a few senior people, and it leaves when they retire.
Video is not enough
Ordinary video loses depth, hand pose and intent.
Lab demos are not real work
Staged tasks miss the gloves, noise and pressure of a real site.
Experts are scarce
The best people are busy and their time is expensive.
How it works
From an expert’s shift to a robot-ready dataset
01 Capture
The expert works as usual wearing an XR headset. We record hands, gaze and the scene.
02 Sync
Hand pose, eye tracking, video and depth are aligned on one timeline.
03 Label
Each task is split into steps, tools and outcomes, checked by people.
04 Deliver
Datasets in the formats robot-learning teams use, or a policy tuned to your task.
What we capture
Hands, gaze and context
Hands
Full hand and finger pose at every moment of the task.
Gaze
Where the expert looks before acting. Attention shows a robot what matters in the scene.
Scene
First-person video and depth of the workspace, tools and parts.
Industries
Built for heavy and hazardous work
Oil and gas
Valves, flanges, inspection rounds on upstream and refinery sites.
Mining and metals
Maintenance and handling in dust, heat and vibration.
Chemicals
Work in protective suits where every mistake is costly.
Energy and wind
Repairs at height and in confined spaces.
Who we work with
Two sides of one problem
For robot makers
- Expert data for manipulation tasks
- Custom capture on real industrial sites
- Access to Central Asia’s industrial base
For industrial sites
- Adapt robots to your processes using your own best workers
- Keep the skills of senior experts before they retire
- You own your data
Why Evoprax
XR builders next to the sites that need robots
XR product team
We have shipped VR products with hand interaction on Meta Quest.
Close to the sites
Based in Kazakhstan, next to world-class oil, gas and mining sites and new robot plants.
Ethics first
Informed consent from every worker, anonymised by default, data owned by the client.
Team
Founders
Evgeny Simonov
Co-founder, XR and hardware. Founder of the VR studio Xocus, which ships titles for Meta Quest. Hand interaction and fast XR prototyping.
Peter Lekarev
Co-founder, product and business. 7+ years in product, 5 years at Alfa-Bank on an app with 7M monthly users. Computer Science, University of Ottawa.
Contact
Let’s talk
We are starting first pilots with robot makers and industrial sites. Tell us about your task.
Email: [add address]
