Design and implementation of a no-code web development platform with an ai-supported large language model

This thesis examines the potential of AI-supported language models in no-code software development processes and focuses on the development and evaluation of the ZOA platform, which represents an innovative approach in this field. ZOA is designed as a no-code platform that allows users to create modern web components using natural language inputs. The platform particularly leverages Natural Language Processing (NLP) techniques to generate HTML, CSS, and JavaScript-based outputs. In the initial phase of the research, the advantages and limitations of existing no-code platforms were analyzed, and their impacts on software development processes were detailed. The literature review identified that AI-supported no-code platforms provide significant cost and time advantages while revealing potential areas for improvement, such as user experience and scalability. During the development of ZOA, a Python-based API and a React-powered user interface were integrated. The model was trained with an extensive HTML and CSS dataset compiled using Tailwind CSS and optimized to interpret user inputs and produce dynamic outputs. The diversity of the dataset and its impact on the model's performance were thoroughly analyzed during the training process. In conclusion, the performance of ZOA was evaluated through various test scenarios, demonstrating its user-friendly structure and potential to democratize development processes. The findings highlight the importance of integrating AI and NLP in future no-code platform development processes. Furthermore, this study provides recommendations for improving ZOA's performance through alternative datasets and optimization techniques.