AI-driven cellular immunotherapy: transforming CAR-T engineering and translation

Chimeric antigen receptor T (CAR-T) cell therapy has demonstrated remarkable success in hematologic malignancies but faces persistent challenges in solid tumors and broader clinical translation, including limited target specificity, heterogeneous tumor antigen expression, unpredictable functional outcomes, and complex manufacturing processes. Recent advances in artificial intelligence (AI) are beginning to transform how these challenges are addressed by enabling data-driven design, prediction, and optimization across the entire CAR-T development pipeline. In this review, we examine how modern AI approaches such as machine learning, deep learning, and generative models are reshaping key stages of CAR-T engineering. We first discuss AI-assisted antigen discovery strategies that integrate multi-omics and clinical datasets to identify tumor-specific targets. We then examine AI-enabled engineering of antigen-recognition modules, including computational design and optimization of antibody- and TCR-derived binding domains. Next, we highlight emerging efforts to program CAR architectures and synthetic signaling circuits using AI models. We further review AI-assisted prediction of CAR-T functional performance, therapeutic efficacy, and clinical outcomes. Finally, we discuss the growing role of AI in manufacturing, quality control, and process optimization, including image-based cellular phenotyping and digital monitoring of production pipelines. Together, these advances suggest a shift from empirical CAR-T engineering toward programmable, predictive, and increasingly autonomous design frameworks that may accelerate the development of safer and more effective cellular immunotherapies.