In the rapidly evolving field of Natural Language Processing (NLP), the task of translating natural language queries into SQL queries (Text-to-SQL) has garnered significant attention due to its potential to simplify database interactions for non technical users. This final project, titled Text-to-SQL on Real World Datasets, explores innovative methods to enhance the accuracy and efficiency of Text-to-SQL systems, specifically focusing on real-world databases with complex schemas. The project leverages the Retrieval-Augmented Generation (RAG) technique to improve Text-to-SQL accuracy by integrating external data sources and fine-tuning strategies. A combination of synthetic dataset generation and prompt strategies is employed to enhance the model s performance. The Mondial dataset, known for its complexity and richness in geographic data, serves as the benchmark for evaluating the proposed techniques. The study aims to develop a robust Text-to-SQL framework capable of handling diverse and complex queries, thereby making database interactions more intuitive and accessible. The methodologies, experiments, and findings documented in this report contribute valuable insights to ongoing research in NLP and database management systems