Orgest Xhelili
Machine Learning & Research Engineer
I am a Machine Learning Engineer at Ramblr.ai working on computer vision models for video intelligence & physical AI. Before joining Ramblr, I obtained my M.Sc. in Computer Science at Technical University of Munich where I specialized in Machine Learning.
I really enjoy working at the intersection of research and engineering, and I am always passionate about exploring new deep learning models and ideas (currently mostly obsessed with generative models).
Latest
Publication
arXiv 2026
BARISTA: A Multi-Task Egocentric Benchmark for Compositional Visual Understanding
Patrick Knab, Orgest Xhelili, Inis Buzi, Drago Andres Guggiana Nilo, Mohd Saquib Khan, Lorenz Kolb, Manuel Scherzer, Kerem Yildirir, Christian Bartelt, and Philipp Johannes Schubert
Blog post
June 2025
A Primer on Flow Matching
A practical introduction to flow matching for generative modeling, including continuous normalizing flows, conditional flow matching, and a PyTorch implementation.
Experience
Machine Learning Engineer @ Ramblr.ai
Nov 2023 – Present • Munich, Germany
Working on computer vision models for video understanding & AI for the physical world.
Software Engineer @ Digital Turbine
Oct 2018 – Oct 2023 • Berlin, Germany
Built high-throughput ad delivery backend systems and distributed data pipelines.
Education
M.Sc. Computer Science - Technical University of Munich (2021-2024)
Specialization in Machine Learning
B.Sc. Computer Science - Jacobs University Bremen (2015-2018)
Minor in Mathematics
Publications
BARISTA: A Multi-Task Egocentric Benchmark for Compositional Visual Understanding
Preprint 2026
Patrick Knab, Orgest Xhelili, Inis Buzi, Drago Andres Guggiana Nilo, Mohd Saquib Khan, Lorenz Kolb, Manuel Scherzer, Kerem Yildirir, Christian Bartelt, and Philipp Johannes Schubert
How Transliterations Improve Crosslingual Alignment
COLING 2025
Yihong Liu, Mingyang Wang, Amir Hossein Kargaran, Ayyoob ImaniGooghari, Orgest Xhelili, Haotian Ye, Chunlan Ma, François Yvon, and Hinrich Schütze
Breaking the Script Barrier in Multilingual Pre-Trained Language Models with Transliteration-Based Post-Training Alignment
EMNLP Findings 2024
Orgest Xhelili, Yihong Liu, and Hinrich Schütze
Posts
A Primer on Flow Matching
June 2025 • Technical Tutorial
A comprehensive introduction to flow matching for generative modeling with mathematical derivations and practical implementation notes.