I am a Brain & AI Scientist at University College London, and a Scientific Programmer at Donders Institute. I am interested in how the Human Brain creates Hallucinations and learns Uncertainty, both huge problems in AI Safety. I want to leverage my unique perspective to help overcome these challenges. I currently lead two multi-year experimental brain scanning studies (7T fMRI and MEG) and one computational modelling study (Deep Neural Networks, Bayesian Modelling) to investigate how human neural networks create false beliefs (hallucination) and calculate uncertainty about its own internal states (introspection) - leading to multiple peer-reviewed articles and high-profile talks. I have professional experience evaluating and training LLMs at Google Assistant, and developed PCNportal, a collaborative open-source Machine Learning app for big data brain analysis.

Humans and LLMs are both interrogable black boxes. I want to use my scientific expertise to make LLMs become more accurate in their thinking, through experimental design and computational modelling of their internal representations.

Download my resumé .