Writings
-
Generative recommendation (Part 2)
In this post, I build a metadata-only recommender using LLaMA-3 models fine-tuned on Amazon review data — achieving SoTA results on a single 24GB GPU with no hyperparameter tuning.
-
Generative recommendation (Part 1)
This post introduces sequential recommenders—what they are, their key types, and common challenges. We'll walk through key preprocessing steps, evaluation splits, metrics, and baselines.
-
Improving Minimax Group Fairness in Sequential Recommendation
A deep dive into my paper on “Improving Minimax Group Fairness in Sequential Recommendation”, accepted to ECIR 2025.