Shpresim Sadiku

Shpresim Sadiku

Applied Scientist, Bonial
PhD in ML & Optimization, TU Berlin
MSc in Data Science, TU Munich

About Me

I am an applied scientist at Bonial, where I build large-scale AI-powered recommender systems spanning representation learning, deep retrieval, and scalable inference — from contrastive embedding models and two-tower multi-interest retrieval to multimodal (text + image) retrieval and LLM-as-a-judge re-ranking, served through real-time GPU inference.

Previously, I completed my PhD in Machine Learning and Optimization at the Institute of Mathematics, Technical University of Berlin, under the supervision of Prof. Dr. Sebastian Pokutta. My research developed optimization-based methods for robust and explainable AI - in particular, inverse classification and counterfactual explanations that make deep neural networks more interpretable and more reliable to train.

Prior to my PhD, I earned an MSc in Mathematics in Data Science from the Technical University of Munich and a BSc in Mathematics from the University of Tirana.

Below, you can find a selection of my publications and talks.

Contact