Uljad Berdica

I am a 2022 Rhodes Scholar and PhD student in the CDT-AIMS program at the University of Oxford focusing on Reinforcement Learning, World Models, Robotics and Reasoning & Exploring with LLMs. Working on meaningful AI that works by learning from experience and Bitter Lesson(s), I am supervised by Prof. Jakob Foerster and Prof. Perla Maiolino from the Oxford Robotics Institute.

My most recent papers include methods to make LLMs more effective and diverse through training or finetuning, unification of Offline RL Algorithms to automatically discover new ones, and efficient bioimpedance measurement solutions in Electronics. Currently exploring Bandits, Gradient-Free methods, stochastic processes and automating research with AI. Interning at J.P. Morgan’s AI Research group as an Associate.

Open to research internships in industry labs. I have served as a reviewer for journals like IEEE Robotics and Automation Letters, conferences like AAAI 2026, ICLR 2026, ICML 2025, IEEE RoboSoft 2024 and several workshops. I like clear papers and clean code. I have opinions on art and comedy which could be silly interpolations of others’.