The Type-Directed Program Synthesis and Repair project aims to leverage type-system specifications and property-based testing for automatically synthesizing and repairing programs, thereby enhancing the software development process. Program synthesis typically involves searching a vast space of potential programs to identify one that meets the given specification. To address this challenge, our proposed approach is to employ a large language model (LLM) as a preconditioner, to generate an initial solution close to the actual one. Subsequently, we will apply program repair techniques such as fault-localization and type-directed synthesis to refine the program and ensure it satisfies the specification. In this project, we will explore the integration of an LLM, such as the Facebook LLaMa model, with an automatic program repair tool, like PropR, to investigate the feasibility and effectiveness of this approach.