About me

I am a PhD researcher in Machine Learning at Imperial College London supervised by Mark van der Wilk. I am also a visiting student at the University of Oxford. Previously, I worked as a research scientist at Babylon Health and was an intern at Amazon AGI Foundations.

Research

I study how machine learning systems can learn inductive biases for predicting well under varying conditions, a useful property for scientific discovery. My work centers on learning causal properties from data and leveraging Bayesian principles (Occam’s razor, calibrated uncertainty) for scalable, data-driven causal modelling. I also explore how insights from deep-learning theory (scaling theory), diffusion and flow-matching methods, meta-learnt foundation models, can amplify these ideas in complex, real-world settings.

I am keen to collaborate on these ideas and in applying these ideas to impactful problems.

A non-exhaustive list of topics I am interested in:

  • Causality: causal discovery, effect estimation, causal representation learning
  • Model selection: Bayesian model selection, MDL
  • Bayesian methods: Bayesian deep learning, Gaussian processes, probabilistic models, inference methods
  • Generative models: Neural processes, Diffusion, flow matching
  • Deep learning: Local learning rules, generalisation, scaling theory, NTK, Tensor Programs, foundation models
  • Information theory: Kolmogorov complexity, information bottleneck

News

  • July 2025: “Estimating Interventional Distributions with Uncertain Causal Graphs through Meta-Learning” accepted at the Scaling Up Intervention Models workshop at ICML 2025. We show that meta-learning enables accurate learning of interventional distributions directly from data, taking model uncertainty (causal structure and mechanism) into account.
  • May 2025: “Continuous Bayesian Model Selection for Multivariate Causal Discovery accepted at ICML 2025. We show that in the multivariate setting Bayesian model selection outperforms other methods in observational causal discovery tasks.
  • March 2025: Won the G-Research Early Career Researcher Grant.
  • Jan 2025: “A Meta-Learning Approach to Bayesian Causal Discovery” accepted at ICLR 2025. We propose a neural process that enables a foundation model like approach for causal discovery, trained entirely on synthetic data.
  • May - Nov 2024: Completed an internship at Amazon AGI Foundations working on scaling neural network architectures while preserving feature learning.