divelab/AIRS

Artificial Intelligence Research for Science (AIRS)

What it solves

AIRS provides a unified, open-source collection of tools, datasets, and benchmarks to accelerate research in AI for Science. It addresses the fragmentation of resources across different scientific domains, specifically focusing on quantum, atomistic, and continuum systems.

How it works

The project organizes AI methods and resources into specialized modules based on the scientific domain:

  • Quantum Mechanics & DFT: Tools like OpenQM and OpenDFT for quantum systems and density functional theory.
  • Atomistic Systems: Tools for small molecules (OpenMol), protein science (OpenProt), materials science (OpenMat), and molecular interactions (OpenMI).
  • Continuum Systems: Tools for solving partial differential equations (OpenPDE) and ordinary differential equations (OpenODE).

Who it’s for

Researchers and scientists working at the intersection of machine learning and physical sciences, particularly those focusing on chemistry, physics, biology, and materials science.

Highlights

  • Comprehensive Scope: Covers everything from quantum-level simulations to continuum-level differential equations.
  • Integrated Ecosystem: Combines software tools, datasets, and benchmarks in one repository.
  • Curation: Based on a comprehensive survey paper detailing AI applications in scientific systems.

Related

  • Project
    RobertTLange/evosax

    evosax is a JAX‑based library that implements 30+ Evolution Strategies (e.g., CMA‑ES, OpenAI‑ES, ARS, SimpleGA) with a vectorised `ask‑eval‑tell` API, enabling fast, hardware‑accelerated black‑box optimisation for neuro‑evolution, RL, and hyper‑parameter search.

  • Project
    pyg-team/pytorch-frame

    A modular deep learning extension for PyTorch designed to build neural network models on heterogeneous tabular data containing mixed column types like text, images, and numerical values.

  • Project
    Junchao-cs/SolarWM

    SolarWM is a unified, open-source framework for training and running long-horizon video world models, supporting multiple backbones like Wan2.2 and MiniMax-H3 with a shared data and training pipeline.

  • Project
    google/ml-compiler-opt

    MLGO is a framework that replaces hand-crafted LLVM compiler heuristics with machine-learned models. It provides training infrastructure and tools for learning optimization policies, currently supporting inlining and register allocation.