Luka Ribar

Luka Ribar is a Research Scientist at Graphcore, focusing on the development of training and inference algorithms for large language models, with an emphasis on efficient techniques that utilize varying forms of sparsity and low precision. Prior to this role, Luka worked as a Research Associate at the University of Cambridge, where contributions included system architecture development aimed at mimicking neural robustness and reliability, alongside dynamical systems modeling and Python software development. Luka also served as a Teaching Associate, supervising students in signal processing, control, and master's research projects, and earlier taught high school students in physics and electronics at Petnica Science Center. Additionally, Luka has experience as a Software Engineer at Microsoft Development Center Serbia, where an optical character recognition system for cursive scripts was developed. Luka earned a PhD in Control and Neuromorphic Engineering and a Master of Engineering in Electrical and Information Sciences from the University of Cambridge.