Physical AI for Materials Intelligence
Physical AI (Robotics · Digital Twin · Recipe Language Model)
for Materials Intelligence
Robotics
Digital Twin
MIRecipe · RLM · MIOS
Materials Intelligence
We introduce a domain-specific RLM developed through seven AI layers and interconnected robotic boxes to drive the evolution of physical AI for defining a new paradigm - Materials Intelligence.
Materials Intelligence Lab Overview
Physical AI System
MIFactory: for Materials Intelligence and Manufacturing
Robotics
The 1st Generation of Robot Platform (2020–2022)
Robotics
The 2nd Generation of Robot Platform (2022–2023)

Robotics
The 3rd Generation of Robot Platform (2024–2026)
Robotics
The 4th Generation of Robot Platform (2026–)

Robotics
The 4th Generation of Robot Platform (2026–)

Robotics
MIT-B-P-F Flexible Robotic Combination for New Battery Materials
Robotics
Engineering Flow
SolidWorks→
BOM→
Electrical Drawing→
Physical Prototype
Digital Twin
Digital Twin Demonstration
Digital Twin
Smart Energy Digital Twin Operation Details
Digital Twin
Prior to Experiment

Digital Twin
During the Experiment

Digital Twin
Post Experiment
MIRecipe
MIRecipe
Reading Literature
60000+papers
Screened Literature
40000+papers
Recipe Resource
20000+files
Recipe Report
12000+reports
Recipe Tokens
578M+tokens
Reading Literature
60000+ papers
Screened Literature
40000+ papers
Recipe Resource
20000+ files
Recipe Report
12000+ reports
Recipe Tokens
578M+ tokens
MIRecipe
MIRecipe
Recipe Language Model
Recipe Language Model
MIOS
MIOS Reading Ability
MIOS
MIOS Doing Ability
MIOS
MIOS System Workflow Overview
Publications
Material Intelligence Framework
Reading-Doing-ThinkingAI + roboticsClosed-loop discovery
Review · Nexus, 2025

Publications
Robotic Nanocrystal Synthesis
Data miningRobotic synthesisInverse design
Article · Nature Synthesis, 2023

Publications
Agentic Robotic Boxes + RLM
11 robotic boxes50,764 PSCs578M+ tokens
Article · Engineering, 2026

Media Wall
Robotics Boxs
Media Wall