Yu Xiang ↗
Co-founder
Research in robotics and computer vision. Professor at UT Dallas; previously NVIDIA, Stanford, and the University of Washington.
INTELLIGENCE BEYOND A SINGLE ROBOT
One intelligence.
Every embodiment.
Allhands Robotics is building the unified data and action layer for robot foundation models.
We are building unified action representations, data, and foundation models that enable manipulation skills to transfer across different robots.
Shared manipulation skills across different hands.Explore the technology ↓
01 / THE PROBLEM
Today’s robots
learn in isolation.
Different robots use different action spaces, embodiments, and datasets. A skill learned on one robot cannot easily transfer to another, making robot data fragmented and expensive to scale.
Robot A→Dataset A→Model A
Robot B→Dataset B→Model B
Robot C→Dataset C→Model C
Separate hardware. Separate data. Separate learning.
02 / THE TECHNOLOGY
One action space.
Many embodiments.
Our unified action representation provides a common interface across different robot embodiments, enabling skills and data to transfer between them.
A SHARED INTERFACEUnified Action Space
Allegro HandLEAP HandShadow HandWUJI HandSharpa Hand
Our technology builds on UHAS, Unified Hand Action Space ↗, our research on learning shared manipulation skills across substantially different robotic hands.
03 / DATA + FOUNDATION MODELS
Turn fragmented data
into shared intelligence.
We are building toward a unified action layer that retargets human and robot data across embodiments—and foundation models that learn from many robots and deploy across many robots.
Human DataRobot DataSimulation Data
Unified Action Space
Cross-Embodiment Data
Foundation Model
Robot ARobot BRobot C
04 / TEAM + CONTACT
Built by people
who build robot intelligence.
We bring together research in robotics, computer vision, and cross-embodiment manipulation to build intelligence that travels.
Luis Felipe Casas ↗
Co-founder
Research in cross-embodiment robot manipulation. Co-author of Unified Hand Action Space.
Jikai Wang ↗
Co-founder
Research in robot perception and dexterous manipulation. Co-author of HO-Cap, a capture system and dataset for hand-object interaction.
We’re building the foundation
for cross-embodiment robot intelligence.
Interested in working with us, partnering, or learning more?
contact@allhandsrobotics.com