Information Retrieval & Machine Learning
Building effective, efficient neural search systems through NLP and deep learning.
I’m Luís Pedro Borges, a Ph.D. student (in progress) at Instituto Superior Técnico / INESC-ID and Carnegie Mellon University. My research connects machine learning models with search systems, with a focus on learned sparse retrieval.
Open to jobs and internships in Canada and Europe: Research Scientist, Data Scientist, Data Engineer, ML Engineer and Search Engineer roles.
Canadian work authorization: IEC Working Holiday approved for up to 2 years; work permit activation on arrival. No employer sponsorship required during the permit period.
Selected Work
KALE: sparse lexical expansion
Uses a K-sparse projector to control the number of active terms in learned sparse representations, connecting neural retrieval with efficient inverted index search.
Read about KALEFinding what users can’t quite name
Combines retrieval and LLM re-ranking for Tip-of-the-Tongue search: finding a known item from an incomplete or imperfect description.
Explore the paper and codeModel ensembles for neural re-ranking
Studies how model ensembling, rank fusion and learning-to-rank improve neural passage re-ranking on MS MARCO.
Read about the studyCurrent Research
My doctoral research, in progress, explores Latent Vocabularies for Learned Sparse Retrieval: neural retrievers that learn their own sparse vocabularies while retaining the efficiency of inverted index search. It brings together sparse representations, scalable knowledge distillation, and control of retrieval cost and latency.
My broader work spans sparse and dense retrieval, passage re-ranking, model ensembles and LLM-based re-ranking, with research presented at ECIR, ICTIR, SIGIR and TREC.
Contact
For job and internship opportunities in Canada and Europe, contact me at lpborges.careers@gmail.com. I’m based in Lisbon, Portugal.