Ali Vardasbi

Ali Vardasbi

Ali Vardasbi

As of December 2023, I have started my new position as a research scientist at Spotify.

My research interests and experiences have spanned a wide range of topics, including information retrieval, bias and fairness, theoretical machine learning, and textual sequence to sequence modeling. A unifying thread among my research endeavors is the pursuit of a theoretical understanding of learning problems. Currently, like many in the field, I am more focused on understanding and utilizing large language models, particularly RAG (Retrieval-Augmented Generation) and ICL (In-Context Learning), for various tasks.

Notes

Here I keep some highlights from the papers I read.

PhD

I joined the University of Amsterdam, IRLab as a Ph.D. on May 2019, under the supervision of Prof. Maarten de Rijke and Dr. Mostafa Dehghani. Our work on fair and unbiased learning to rank (LTR) from user interactions has led to several papers in SIGIR and CIKM. We also explored overparameterization and memorization/generalization in general machine learning.

In the summer of 2022, I interned in the machine translation team at Apple and extended my experience in sequence modeling, particularly transformers and state space models.

In the final year of my PhD (i.e., 2023), I have worked on generative Large Language Models (LLMs) and retrieval augmentation techniques in music recommendation as an intern at Spotify.

Thesis

Re-Examining Assumptions in Fair and Unbiased Learning to Rank. Amsterdam, December 2023.

Publications

Conferences

Journals