At University of A Coruña, I published 8 first-author papers on LLM evaluation, data-efficient NLP, structured prediction, and multilingual transfer.
I developed sequence-labeling methods for parsing and nested named entity recognition. I also built reusable PyTorch and Hugging Face training and evaluation frameworks, including analysis pipelines for human- and LLM-generated text.
At EPFL, I studied how transformers represent and generalize knowledge in Antoine Bosselut’s NLP Lab. I built Docker and Run:ai workflows to queue, monitor, and log parallel experiments across multiple A100 GPUs.
At Alstom, I replaced a manual spreadsheet process with a Python and SQLite application, adding validation and reporting used by engineering stakeholders.

