LMGenDrive: Bridging Multimodal Understanding and Generative World Modeling for End-to-End Driving | ScienceToStartup