Expert data for physical AI
Experts know how hard to push and when to stop. We record it, so robots can learn it.
Watching an expert
Your turn: press and drag down Hold Space to push. Up arrow to ease off.
Force
0.0 newtons
Touch: what you feel
PASS
An expert, like a surgeon, works by feel: the part a camera cannot see.
We record what they feel, how hard they push, what they see and what they say.
A robot learns from those recordings.
Then the robot is tested against what experts did. You set the pass rule.
Expert force + touch data to train and test robots.
Embodied AI learns from our recordings, then is tested against them.
Every clinician we record holds a current license. They work on practice models and animal tissue, not patients.
Force is recorded in newtons, the same unit a robot's own sensor reports.
You set the requirements. You pay only for recordings that meet them.
For embodied AI that works on soft tissue, where touch matters more than sight: surgical robots and tissue simulation.
Research
Our dataset paper: an expert surgeon's force, video and spoken reasoning, recorded together and packaged for embodied AI in the open LeRobot format. As a proof of concept, we fine-tune the open NVIDIA Isaac GR00T model on it.
Humynex Robotics, Inc. · Beverly Hills, California
Co-Founder · CEO
A practicing cosmetic surgeon in Los Angeles for about four decades, with more than 10,000 liposuction procedures by his own count. He is the surgeon behind our surgical dataset.
Co-Founder · Operations
More than 35 years running outpatient surgery centers. He leads operations: the rooms, the credentials, the consent.
Co-Founder · CTO
A machine learning scientist with published research on benchmark datasets. She built our recording system, the dataset and the link to embodied AI models.
Tell us which soft-tissue task your robot or simulator needs to get right.
Thank you. We will reply by email.