The development of cyber-physical systems presents fundamental challenges in system design, verification, and implementation, particularly in safety-critical domains where reliability and real-time performance are paramount. Classical development approaches rely heavily on manual coding and modeling, which are resource- intensive, error-prone, and require substantial domain expertise. While Model-Driven Engineering (MDE) methodologies, such as the Architecture Analysis and Design Language (AADL), provide systematic abstraction tools, they struggle with the dynamic and unpredictable nature of CPS environments. To address these challenges, we propose an integrated method that combines AADL automation with code generation using large language models (LLMs) to automate model generation, intelligent code synthesis, and error correction. Our approach uses a dual-phase approach: first, automating AADL model generation from natural language requirements using domain-specific LLMs, and second, implementing intelligent code generation with Chain-of- Thought prompting for automated error detection and correction. The framework integrates formal verification techniques, runtime assurance mechanisms, and version control to ensure safety-critical compliance while reducing development time and manual effort. We validate our approach through comprehensive case studies including unmanned aerial vehicles and smart building systems, showing improvements in development efficiency, code quality, and system reliability. Our results show an average 70% reduction in development time, syntax accuracy exceeding 88%, and a 25% reduction in energy consumption for deployed systems.