Description
We recently submitted a research paper to the IEEE Transactions on Dependable and Secure Computing (TDSC) journal, which explores the integration of N-version programming with automatic programming language translation, utilizing large language models (LLMs). Additionally, I undertook a complete refactoring of a tool, developed in both Python and C++, designed to aid in the development and testing of LLM-driven automatic code translation, for the aforementioned paper.
Currently, my focus is on developing an automated feedback system aimed at addressing software supply chain errors commonly made by students, using dirty-waters as its driver. Toward this, I have had to focus on this both from a research assistant and as a Master’s thesis student perspective:
- From the former, I have been responsible for implementing Maven support for dirty-waters, as well as bug fixing, refactoring, documenting, and implementing caching features – the latter of which has been particularly challenging but rewarding, with the system now being able to run in around 5-10s, down from 10-15 minutes for the Spoon project, as an example.
- From the latter, I have been responsible for the creation of a CI action for dirty-waters, in order to be able to use the tool in an easier manner, both in student and developer environments. I will then proceed to evaluate the tool in real-world scenarios, by being employed both in open-source projects as well as in courses at KTH.