MsbA, a bacterial ABC transporter essential for lipopolysaccharide (LPS) transport, is a promising target for treating multidrug-resistant infections. However, developing potent MsbA inhibitors remains challenging. Recently, cerastecin C, representing a class of bivalent inhibitors targeting the MsbA dimer interface, demonstrated in vivo efficacy against multidrug-resistant Gram-negative bacteria, despite high plasma protein binding, by blocking LPS translocation essential for bacterial survival. In this study, we developed a bivalent molecular generative model, BL-INVENT, focusing on linker design to optimize the structure of cerastecin C. Several generative model designed compounds were synthesized and exhibited antibacterial activity. Further optimization led to the identification of compound 12, which showed enhanced potency and lower plasma protein binding compared with cerastecin C. These findings demonstrate an effective AI-driven strategy for the development of next-generation MsbA inhibitors.