Tools, Data & Consulting Services for A.I.-Driven Materials Design

Tools

Automated machine learning modeling
Analysis of data structure and features
Optimization of processing procedures
Screening millions of candidate materials

Data

Materials data repositories
Custom data ingestion & scraping
Management services
Data infrastructure

Consulting

Domain expertise in materials modeling
Highly flexible project structure
Machine learning training
Build internal machine learning efforts

Book meeting

Schedule a 15-minute introductory consultation with the Aionics team to assess the applicability of materials informatics to your R&D.

Mateo Jaramillo

CEO and Co-Founder, Form Energy

The Aionics platform enables us to seamlessly implement machine learning models on our in-house data in real time, which brings another layer of rigorous analysis to help identify the chemical features driving performance that we care about most. The automated machine learning model-building and screening workflow also lets us screen hundreds of thousands of candidate materials for our devices in the blink of an eye.

A.I. Approaches Offer Significant Acceleration

Project Structure

Phase 1

Pipeline construction

0-6 months consulting

Phase 2.1

Deployment & Training

1 week

Phase 2.2

Potential Consulting

Development as requested

Phase 3

Supported Deployment

MATLAB-like licensing

We seek to mitigate risk by combining high and low risk deliverables.

Machine learning-based performance predictions are considered high risk;

low risk deliverables include data management, candidate material scraping, historical data analysis.

Our customers & partners

Platform: Data overview

Data assets:

Materials data repositories
A.I. data aggregation: OCR, NLP, etc.
Human data entry
Data management and infrastructure
Custom data integrations
Close connections with academic materials ML community

Existing datasets:

Ceramics: 80,000+

80,000+ crystal structures, formation energies, band gaps, estimated materials costs
18,000+ Li materials: predicted ionic conductivities, thermodynamic redox potentials
1,500+ Li cathodes: voltages, capacities, electrochemical expansions
100+ known Li conductors: ionic conductivities

Molecules: 150,000+

150,000+ molecular structures with 10+ physical properties, supplier information/price, toxicity, flammability
5,000+ with known redox potentials vs. Li
1,000+ salts

Polymers: ~1,000

~10 experimental properties and ~10 computed properties
Existing capabilities to model mixed (hybrid) solutions

Trace Additives: infinite

10+ predicted properties from ML models w/o structure