Numerical simulations in engineering and science are often costly and time-consuming. While machine learning-based surrogate models offer a solution, they typically require large datasets. This project aims to develop adaptive models that can be repurposed for similar tasks, such as new PDE problems, using minimal data. By leveraging sparse data and techniques like transfer learning, meta-learning and few-shot learning, the project will enhance the efficiency and accuracy of these models, reducing the need for large datasets and enabling faster assessments.