On-demand design of controlled-release systems using an expert-mimic AI framework

Navigating a vast formulation-preparation design space that is highly sensitive to perturbations makes trial-and-error approaches inefficient and costly for diverse clinical needs. Here, we introduce an expert-mimic AI framework called E-MAF that integrates a process-aware surrogate with an optimized genetic algorithm to emulate coarse-to-fine reasoning, autonomously generating lab-ready designs tailored to desired release profiles. We show that E-MAF delivers on-demand release and shortens design cycles from 6-12 months to approximately 1 hour. We demonstrate that in vitro release, in vivo pharmacokinetics and therapeutic effects match or surpass commercial benchmarks across 7 active pharmaceutical ingredients. We show that E-MAF also enables release profiles that are hard to achieve by conventional approaches, exemplified by ≤2% risperidone burst within 1 h and near-zero-order release over ~28 days in vivo. These results demonstrate E-MAF’s potential to accelerate polymeric controlled-release system development and drug-delivery industrialization. Here the authors develop an AI framework called E-MAF that, starting from a desired release profile, couples process-aware prediction with expert-mimic evolutionary search to generate interpretable, lab-ready formulations for controlled release systems.