Human–AI Co-Embodied Intelligence

APEX: The Agentic-Physical Experimentation System

Mixed-reality goggles, agentic AI, and human researchers united in a closed loop — bringing real-time reasoning, guidance, and protocol co-development into the microelectromechanical systems (MEMS) cleanroom.

Demo video coming soon

About APEX

APEX is a physical AI framework in which agentic AI provides contextual reasoning, adaptive planning, and real-time analysis, while humans perform dexterous physical execution. Wearable mixed-reality hardware streams experimental context to the AI and delivers in-situ guidance back to the operator.

Human–AI Co-Embodied Intelligence for MEMS Fabrication

Outer Loop: Real-time Protocol Co-Development

Inner Loop: Real-time MEMS Device Co-Fabrication

System Workflow

Click any node to see what it does. This diagram will grow into the full interactive workflow explorer.

Human Operator MR Goggles Perception + Display Local Agents Safety · Wafer Check High-Level Agents Plan · Context · Track · Analyze Evolving Memory Logs · Experience Guidance Back to human

Paper & Team

Human–AI Co-Embodied Intelligence for MEMS Scientific Experimentation and Manufacturing.
Manuscript in preparation — link coming soon.

Xinyi Lin*, Yuyang Zhang*, Yuanhang Gan, Juntao Chen, Hao Shen, Camilla Meneghetti, Yichun He, Lijun Li, Ze Yuan, Shuang Wang, Chaohao Wang, Ian Van Hulle, Rui Zhang, Na Li, Jia Liu

John A. Paulson School of Engineering and Applied Sciences, Harvard University · UltraReality Technology Limited
* equal contribution · † corresponding: nali@seas.harvard.edu, jia_liu@seas.harvard.edu