LEVIATAN – Leveraging Open-source AI for Embedded Software Development

Abstract

This project aims to implement a large language model (LLM) agentic solution tailored for embedded code development. By utilizing open-source, self-hosted models, it ensures the use of confidential data without privacy concerns. Additionally, an automatic pipeline has been developed to facilitate the computation of metrics, assessing the performance of the solution and enabling its benchmarking with other solutions on the market. Ultimately, this project improves embedded software development engineers' productivity by focusing their effort on the most meaningful tasks.