← All companies

Science / Company profile

Radical AI

What Radical AI does

Radical AI is an AI-for-science company focused on accelerating materials discovery by tightly integrating (1) AI that proposes materials and experiments, (2) simulation and learning components, and (3) a self-driving/robotic laboratory that executes experiments and feeds measured data back into the system. The company positions its approach as “closed-loop” scientific intelligence: scientists provide performance targets, AI-generated experiments run in the lab, and results are iteratively refined until goals are met.

Technically, Radical AI has published and open-sourced parts of its underlying stack, including TorchSim, a PyTorch-native atomistic simulation engine intended to speed atomistic workflows and support real-time correlation with lab measurements. In addition, it has described ML-centric simulation capabilities such as EGIP (a direct-force interatomic potential) as well as research tooling for extracting experimental “process + measurement” detail from scientific literature (LitXBench).

Commercially, the company states that it will “sell materials” to industries including energy, aerospace, defense, semiconductors, and automotive, and it has pursued U.S. defense R&D funding via AFWERX (Direct-to-Phase II SBIR/STTR). It also announced and expanded a major footprint in New York tied to building autonomous materials science labs at the Brooklyn Navy Yard, with government-backed tax credits and an expectation of high throughput of AI-driven experiments per day.

News

Company record

Aug 07, 2026 · Official · Radical AIPurdue Applied Research Institute Confirms Superior Performance of Radical AI Materials

Radical AI says Purdue verified that a Radical AI–invented material (RAI-939) outperformed industry aerospace material C103 in a head-to-head torch test, reporting markedly different mass changes after a minute of testing, and describing plans to scale production for larger test articles.

Jun 09, 2026 · Official · Radical AIAutomated Microstructure Analysis Moves Different

Radical AI describes an automated SEM-image analysis workflow for estimating secondary dendrite arm spacing (SDAS), reporting that it has been collecting 10,000+ SEM images and that the automated pipeline takes about a second per image on a personal computer to produce structured, comparable data for annealing decisions.

Aug 25, 2025 · Official · Radical AIRadical AI Awarded U.S. Air Force Contract

Radical AI announced it was selected by AFWERX for a Direct-to-Phase II contract worth $1,197,902 focused on accelerated discovery of novel high entropy alloys (HEAs) for the Department of the Air Force, describing use of AI-driven predictions, high-throughput computational screening, and robotically operated labs.

Jul 18, 2025 · Official · Radical AIRadical AI announces $55M in new funding

Radical AI announced the close of a $55M Seed+ round to accelerate its mission to autonomously discover, create, and manufacture next-generation materials using an integrated, closed-loop approach.

Apr 24, 2025 · Official · Radical AIEGIP: Another step forward in ML-first materials

Radical AI announced EGIP, describing it as a first-generation interatomic potential intended to confirm the performance and speed benefits of direct force prediction, including claims of efficiency improvements and accuracy on properties such as thermal conductivity and crystal structure stability.

Apr 04, 2025 · Official · Radical AIIntroducing TorchSim

Radical AI announced release of TorchSim, describing it as a PyTorch-native atomistic simulation engine for the MLIP era and stating it is open-source, positioned as a component in a broader materials flywheel intended to connect simulation with lab measurements.

Source map · 2 recurring channels · 16 references

Still resolving: Newsroom · Official X

What it builds

TorchSim

TorchSim is a PyTorch-native atomistic simulation engine released by Radical AI, intended to enable faster atomistic simulation workflows in support of its self-driving lab approach.

source ↗
EGIP

Efficient Geometric Interatomic Potential (EGIP) is Radical AI’s first-generation interatomic potential, presented as an ML-first, direct-force approach designed to improve the speed/accuracy trade-off for atomistic simulations.

source ↗
Self-driving / autonomous materials lab

Radical AI’s self-driving lab executes AI-generated experiments and iteratively refines results based on measured performance, with the company publicly describing a closed-loop workflow connecting prompts to experimental execution.

source ↗
Sentinel (lab informatics platform)

Sentinel is described by Radical AI in its 2025 recap as a core lab informatics platform tied to its autonomous laboratory stack.

source ↗
Antimatter (LabOS)

Antimatter (LabOS) is described by Radical AI in its 2025 recap as part of its autonomous lab software stack, shipped in multiple versions.

source ↗

Milestones & partnerships