The aim of the thesis is to implement a large language model (LLM)-based pipeline for program synthesis of Python program with iterative refinement based on unit tests. The user will provide a textual description of the functionality of the function to be synthesized as well as unit tests for that functionality. The program output by the LLM will be tested using the provided tests and the test results will be passed back to the LLM with a prompt asking for program repair. The thesis will also include an experimental evaluation of the pipeline w.r.t. various free coding-focused LLMs, classes of problems, and prompts. The text part of the thesis will briefly provide the necessary background. Then, it will focus on describing the design choices and implementation behind the proposed pipeline and thorough presentation of experimental outcomes.