aims-PAX: Parallel Active eXploration for the automated construction of Machine Learning Force Fields

Recent advances in machine learning force fields (MLFF) have significantly extended the capabilities of atomistic simulations. This progress highlights the critical need for reliable reference datasets, accurate MLFFs, and, crucially, efficient active learning strategies to enable robust modeling of complex chemical and materials systems. Here, we introduce AIMS-PAX, an automated, multi-trajectory active learning framework that streamlines the development of MLFFs. Designed for both experts and newcomers, AIMS-PAX offers a modular, high-performance workflow that cou- ples flexible sampling with scalable training across CPU and GPU architectures. Built on the widely adopted ab initio code FHI-AIMS, the framework seamlessly integrates with state-of-the-art ML models and supports pretraining using general-purpose (or "foundational") models for rapid deployment in diverse systems. We demonstrate the capabilities of AIMS-PAX on two challenging systems: a highly flexible peptide and bulk CsPbI3 perovskite. Across these cases, AIMS-PAX achieves a reduction of up to two orders of magnitude in the number of required reference calculations and enables over 20x speedup in AL cycle time through optimized resource utilization. This positions AIMS-PAX as a powerful and versatile platform for next-generation ML-driven atomistic simulations in both academic and industrial settings.