Mat3ra
The materials R&D cloud.
The company I founded in 2015 and lead today. It exists because the simulation codes that decide what a material will do are exact, powerful, and miserable to run at any scale.
What it does
Design a structure, run the simulation, and build models on the results, without standing up high-performance computing infrastructure first. Density-functional theory, molecular dynamics, and machine-learned force fields run through one workflow engine and API.
How it got here
Founded as Exabyte Inc.
Started out of Alchemist Accelerator, on the premise that materials simulation belonged in the cloud rather than on a departmental cluster.
Backed by Breakout Labs
A grant from the Thiel Foundation's fund for early, science-heavy companies, alongside published benchmarks showing public clouds could match traditional supercomputers on Linpack.
Building for federal laboratories
Began a run of research contracts and grants with the U.S. Air Force, the Department of Energy, and NIST, developing materials modelling infrastructure and data standards.
Became Mat3ra, and raised a seed round
Three million dollars led by Draper Associates, with Albert Wenger and Serg Bell, and a new name for the platform.
AI-ready materials data
Published Mat3ra-2D with a co-author at NIST: an open framework for generating realistic two-dimensional materials and interfaces for machine-learning datasets.
Who uses it
- Shell
- U.S. Air Force
- U.S. Department of Energy
- NIST
- Tokyo University of Science
- Tohoku University
- ITOCHU Techno-Solutions
- Intermolecular, acquired by Merck KGaA
Released openly
The libraries underneath the platform are open source, so the data and the schemas outlive any one vendor—including ours.
In the press
Startup That Models Materials at Atomic Level Raises Seed Capital
CIO Journal, by Suman Bhattacharyya, on the seed round led by Draper Associates.
Azure Edges AWS in Linpack Benchmark Study
On the study measuring whether public clouds could hold their own against traditional supercomputers.
Clouds can compete with HPC, say boffins
Coverage of the same benchmarking work, run against NERSC's Edison supercomputer.
Exabyte Measures Linpack Performance Across Major Cloud Vendors
The supercomputing list's own write-up of the benchmark results.
Media resources
Mat3ra (San Francisco, founded 2015 as Exabyte Inc.) develops a cloud platform for AI-driven materials R&D: atomistic design, simulation, and machine learning in one environment. The company is backed by Breakout Labs and a seed round led by Tim Draper, has received $3,888,133 across eight competitive federal R&D awards, and its platform is used by Shell, the U.S. Air Force, the U.S. Department of Energy, Tokyo University of Science, and Tohoku University.