This doctoral project will focus on the development of interpretability and watermarking techniques for large language models (LLMs). It will address both methods to unravel the internal logic of transformers and strategies to embed digital watermarks into generated responses without degrading quality. The research will provide insights into:
Through deep learning techniques, statistical analysis, and empirical testing, the selected candidate will join a research team committed to advancing transparency and accountability in the use and deployment of LLMs.
Master’s degree in Engineering, Computer Science, Physics, or Mathematics. Candidates with a Master’s in Big Data will be especially valued.
High level of spoken and written English and Spanish.
Strong programming skills in Python, with a particular focus on working with large language models (LLMs) and vision-language models (VLMs). Prior experience with watermarking techniques applied to LLMs is highly valued.
Strong teamwork skills. Research project experience is an asset.
This position will enable the predoctoral researcher to establish a close professional relationship with leading companies and institutions, thereby enriching his/her research with the practical experience gained through cooperative projects.
Initially, this arrangement does not necessarily require the predoctoral researcher to enrol directly in a Doctoral program. Admission to a Doctoral program can be requested within the first two years.
The predoctoral researcher is hired on a yearly basis for a maximum period of six years. However, if the researcher is not enrolled in a Doctoral program, the contract cannot exceed two years.
The gross salary for the first year is approximately €31.472,55 (2025). Additionally, the IIT covers 90% of the fees for training activities, and the researcher will have access to all university services, including sports, cultural activities, and the library.
1
