Hi! I’m Alex, a Machine Learning PhD student at Cambridge, currently interning at Spotify. I am interested in developing safe and robust machine learning systems that behave in the way we expect, and thus able to work with humans effectively. My work has often explored inverse reinforcement learning and imitation learning in order to learn from humans, and I’m interested in using this knowledge and generative AI to create insight and develop personalised, human-centric, decision making systems.
If for some reason you’re interested in who I am or what I do then you’ve come to the right place. However, if you can’t find something feel free to get in contact through one of the many ways available in the modern world - including my email and this form.
CV | Email | Google Scholar | Github | LinkedIn | Twitter
How to Catch an AI Liar: Lie Detection in Black-Box LLMs by Asking Unrelated Questions accepted to ICLR 2024!
Two papers accepted to NeurIPS 2023: AllSim: Systematic Simulation and Benchmarking of Repeated Resource Allocation Policies in Multi-User Systems with Varying Resources and GAUCHE: A Library for Gaussian Processes in Chemistry! Also a fun workshop paper: Optimising Human-AI Collaboration by Finding Convincing Explanations!
Starting as a Research Scientist Intern at Spotify as part of the Satisfaction, Interaction and Algorithms (SIA) team, working on intergratings large language models into the content moderation pipeline using culturally finetuned models.
Taking part as a SERI-MATS Scholar working with Owain Evans, investigating deception in large language models along with methods to detect a notion of lying using only black-box access.
Accepted to NeurIPS 2022: Synthetic Model Combination: An Instance-wise Approach to Unsupervised Ensemble Learning!
Awarded second place in the G-Research PhD prize in maths and data science for best PhD draft dissertation!
Again, two papers accepted to ICLR! Inverse Online Learning: Understanding Non-Stationary and Reactionary Policies and POETREE: Interpretable Policy Learning with Adaptive Decision Trees.
The Medkit-learn(ing) Environment: Medical Decision Modelling through Simulation is accepted to the Neural Information Processing Systems (NeurIPS) main conference track on Datasets and Benchmarks.
First place in the Hex Cambridge hackathon Optiver Challenge out of about 40 teams. Developed an algorithmic trading strategy for market making a dual listed product.
Two papers accepted to ICLR! Scalable Bayesian Inverse Reinforcement Learning and Generative Time-series Modeling with Fourier Flows.
Graduated the MPhil in Machine Leaning and Machine Intelligence and started a PhD in Machine Learning for Healthcare at the University of Cambridge.
My first ever paper accepted to ICML: Unlabelled Data Improves Bayesian Uncertainty Calibration under Covariate Shift.