Overview
FHI-aims (Fritz Haber Institute ab initio molecular simulations) is a leading all-electron electronic structure code based on numeric atom-centered orbitals (NAO). It is designed for high-performance, large-scale simulations of molecules and materials. In the 2026 landscape of AI for Science (AI4S), FHI-aims occupies a critical position as the primary high-fidelity data generator for training Machine Learning Interatomic Potentials (MLIPs) and Neural Network Force Fields. Unlike pseudopotential-based codes, FHI-aims treats all electrons explicitly, providing a benchmark-level accuracy that is essential for fine-tuning foundation models in materials discovery. The architecture is highly parallelized, optimized for exascale computing, and features a robust integration with the Atomic Simulation Environment (ASE). Its capability to handle hybrid functionals, many-body perturbation theory (GW, MP2), and relativistic effects across the periodic table makes it the preferred tool for researchers moving from traditional simulation to AI-integrated materials engineering workflows.