Learning Matter @ MIT

Intergrating Atomistic Skills into Agentic IDEs (Cursor, Claude Code, Google Antigravity, OpenClaw, etc)

Graph Neural Network to predict the reaction related properties for reinforcement learning

Neural Network Force Field based on PyTorch

Atomistic Language Models unify natural language and atomistics to understand, generate, and optimize materials. https://arxiv.org/abs/2606.21395

Reinforcement learning driven simulation of vacancy diffusion

A platform for setting up high throughput polymer electrolyte MD simulations.

Mechanism-based reaction-diffusion model to predict FROMP front propagation from monomer molecular descriptors.

Update LigParGen version from Leela Sriram Dodda

Code for XRD-based structure optimization that was used in the study presented in "The Loss Landscape of Powder X-Ray Diffraction-Based Structure Optimization Is Too Rough for Gradient Descent".

MCMC-based algorithm for sampling surface reconstructions