Overview of Proception
Proception (proception.ai) is a robotics and physical artificial intelligence company headquartered in Palo Alto and Mountain View, California, United States. Founded in 2024 by Jay Li (a former electronics lead for Tesla's Optimus humanoid robot program with hardware engineering experience at Apple, Aeva, and Aurora) and Jack Xu, Proception is backed by Y Combinator and leading venture capital firms.
The company focuses on solving one of the most complex challenges in physical AI: achieving human-level, high-dexterity robotic manipulation. Proception engineers advanced humanoid robotic hands and control systems designed to replicate the subtle articulation, tactile feedback, and force sensitivity of human hands.
Continue…What the Company Does
Traditional industrial robot grippers are limited to rigid, binary actions such as clamping or vacuum suction. Proception builds robotic hardware and software architectures designed for fine motor skills, adaptive grasping, and dynamic object manipulation in unpredictable real-world environments.
By combining tendon-driven mechanical architectures with high-density tactile sensors and real-time AI learning models, Proception allows humanoid systems and collaborative robots (cobots) to handle fragile objects, adjust grips dynamically, reorient items in-hand, and perform intricate manual tasks previously restricted to human workers.
Core Products, Hardware, and Services
1. ProHand (Flagship Humanoid Robotic Hand)
ProHand (including the research-grade ProHand 1.0) is Proception's flagship hardware and sensing system designed for advanced dexterity and research deployment.
Key technical specifications and capabilities include:
* Tendon-Driven Mechanical Architecture: Features more than 20 degrees of freedom (DoF), allowing individual joint flexibility and fluid movement patterns similar to human fingers.
* Integrated Tactile Sensing: Embedded with distributed high-resolution tactile "skin" sensors and pressure arrays that capture contact data, surface texture, and resistance in real time.
* Sub-Millimeter Control & Force Sensitivity: Delivers delicate force regulation and high-precision motor control, enabling fine tasks such as threading needles, holding delicate glassware, or adjusting objects without dropping or crushing them.
* Compliant and Adaptive Grasping: Combines joint position awareness with active physical compliance to automatically conform finger shapes to irregular or moving objects.
2. Physical AI Learning Models & Control Software
Proception develops real-time control algorithms and physics-informed AI models to power object manipulation:
* Real-Time In-Hand Manipulation: AI models process tactile data and spatial telemetry continuously to execute live grip adjustments, object reorientation, and balance correction.
* Simulation and CAD Integration: Utilizes automated CAD-to-URDF (Unified Robot Description Format) pipelines integrated with design platforms like Autodesk Fusion, reducing configuration times for physical AI research environments from hours to seconds.
3. Deployments and Developer Services
- Research-Grade System Shipments: Proception supplies ProHand units directly to artificial intelligence research institutions, enterprise robotics labs, and humanoid robot developers working on general-purpose physical AI.
- Custom Integration and Hardware Engineering: Provides dedicated engineering support to help robotics teams integrate ProHand hardware, tactile feedback loops, and control software into existing humanoid frames, robotic arms, and automated laboratory platforms.
