Ilija Bogunovic

Ilija Bogunovic is an Assistant Professor (UK Lecturer) in the Department of Electronic and Electrical Engineering at University College London (UCL), Faculty of Engineering Sciences. His research focuses on algorithmic sequential decision-making for robust, reliable, and safe artificial intelligence, with applications in large language model alignment, reinforcement learning, and human feedback integration.

His research interests lie at the intersection of machine learning, optimization, and decision theory. He investigates robustness in Bayesian optimization, bandits, and reinforcement learning under adversarial attacks, model misspecification, and distributional shifts. His work spans both theoretical foundations and real-world applications in energy systems, mobility, and AI safety. A central theme is developing algorithms that are provably robust and efficient in uncertain and potentially corrupted environments.

The recent publications highlight a strong trend in robust decision-making, particularly in reinforcement learning and preference optimization. Key themes include adversarial robustness in large models, distributional robustness in RL, safe multi-agent systems, and robust Bayesian optimization. The work frequently involves theoretical analysis of regret and sample complexity, combined with practical algorithm design for high-impact applications.

Scientific Awards:

• Google Research Scholar Program Award (Machine Learning & Data Mining), 2023
• EPSRC New Investigator Award, 2023

Ilija Bogunovic actively supervises PhD and MSc students, with several of his advisees leading papers accepted to top venues like NeurIPS and ICLR. He has secured competitive research grants, including the EPSRC New Investigator Award, to support his work on robust decision-making. He is building a research team focused on robust and safe AI.

He leads a research team and has initiated an online reading group on modern adaptive experimental design and active learning, fostering a collaborative research environment.