AtomisticSkills: AI Framework for Atomistic Research
AtomisticSkills, a new open-source framework, is designed to improve general-purpose AI coding agents specifically for atomistic research fields, including materials science, chemistry, and drug discovery. The framework organizes scientific workflows into modular components, offering more than 100 specialized skills and tools. Among its capabilities are database access, thermodynamic and kinetic modeling, and the utilization of simulation engines leveraging machine learning interatomic potentials and density functional theory. This innovative method aims to overcome the difficulties in scaling larger agents to meet the intricate demands of atomistic research across various software environments.
Key facts
- AtomisticSkills is an open-source framework.
- It empowers general-purpose AI coding agents for atomistic research.
- The framework hierarchically decomposes scientific workflows into agent skills and tools.
- It integrates over 100 human-curated multidisciplinary skills.
- Skills include database access, thermodynamics and kinetics modeling, and simulation engines.
- Simulation engines employ MLIPs and DFT.
- The framework targets materials science, chemistry, and drug discovery.
- It addresses challenges of scaling monolithic agents in atomistic research.
Entities
—