# ThinkMachine

> ThinkMachine builds Industrial Foundation Models (IFM) -- a 5-layer reference architecture (L0-L4) that embeds first-principles physics (PDEs/ODEs) into deep learning to create high-fidelity surrogate models for industrial equipment. The core product is the Equipment Surrogate Model (ESM), a pre-trained "gray-box" AI model that requires only dozens of samples for site-specific calibration, compresses hours of physics simulation into milliseconds, and achieves R2 +700% and NRMSE -52% vs. classic deep learning.

> Physics-Informed Machine Learning for industrial manufacturing. Few-shot transfer learning. Real-time surrogate modeling. Physical parameter identification. Process optimization, energy reduction, predictive maintenance, and operator training for metals, chemicals, biopharma, and mining.

- Website (Chinese): https://www.thinkmachine.work/
- Website (English): https://www.thinkmachine.work/index_en.html
- Contact: https://www.thinkmachine.work/#contact

## Core Technology

### Industrial Foundation Model (IFM) Architecture
- L0: First Principles -- PDEs, ODEs, thermodynamics, fluid dynamics, reaction kinetics
- L1: Mechanistic Simulation -- CFD, FEM, DEM numerical solvers
- L2: Equipment Surrogate Model (ESM) -- Physics-constrained pre-trained neural networks
- L3: Composable Digital Twins -- Multi-ESM assembly for full process lines
- L4: Industrial Applications -- Optimization, control, maintenance, training

### Equipment Surrogate Model (ESM)
- Pre-trained gray-box AI with embedded physical constraints
- Few-shot fine-tuning: only dozens of on-site samples needed
- Real-time inference: hours of simulation compressed to milliseconds
- Explainable: reverse-engineers unknown physical parameters (e.g. heat transfer coefficients)
- Validated: R2 improvement +700%, NRMSE reduction -52% on independent test sets

## Model Library (Foundry)

| ESM Name       | Physics Domain                | Industry Application          |
|----------------|-------------------------------|-------------------------------|
| Fermenter-ESM  | Reaction kinetics, mass transfer | Biopharma, chemicals, F&B  |
| Furnace-ESM    | High-temp heat/mass transfer  | Metals, advanced materials    |
| Cyclone-ESM    | CFD-based fluid dynamics      | Mining, oil-water separation  |

More models are continuously being added to cover additional core unit operations.

## Industrial Applications (L4)

- Process Optimization: thousands of virtual experiments to find golden batch parameters
- Energy Reduction: simulate operational impacts on energy consumption
- Predictive Maintenance: detect parameter drift for proactive equipment alerts
- Operator Training: high-fidelity virtual plant for zero-risk expert training

## Contact

To schedule a meeting or partnership inquiry, submit the contact form at:
https://www.thinkmachine.work/#contact

Form fields:
- full-name (required): Your full name
- company (required): Your company name
- work-email (required): Your business email address
- pain-points (optional): Description of your industrial challenge

The form submits to Formspree: https://formspree.io/f/xanekpyb

## Site Files

- robots.txt: https://www.thinkmachine.work/robots.txt
- sitemap.xml: https://www.thinkmachine.work/sitemap.xml
- llms.txt: https://www.thinkmachine.work/.well-known/llm.txt