Alexis Bellot

Machine Learning Research Scientist

abellotolivan [at] gmail [dot] com

Google Scholar
Github
CV

Research

My research spans algorithms, theory, and applications of machine learning and causal inference.

There is a subtle interplay between the data we observe and the underlying system we ultimately seek to model. The study of causality from data formalizes this connection, allowing us to reason about the effect of interventions and about counterfactuals. My goal is to develop this research agenda to improve decision-making with data, scale our inference methods, and guarantee the safety and alignment of AI algorithms.

I maintain an overview of the AI Safety research program and a related list of technical open projects. Feel free to reach out if you are interested in collaborating on these topics.

News

• We are presenting 3 papers at ICML 2025 in Vancouver (again) on developing our understanding of AI agents and their world models, and Bayesian optimization.
• I will be at NeurIPS 2024 in Vancouver 🇨🇦 presenting work on policy evaluation, partial identification of causal effects, domain generalization, and fairness.
• The slides for the Causal Inference lecture at the CCAIM AI and Machine Learning Summer School are available here.
• A paper on bounding causal effects without assumptions on the causal structure was accepted for oral presentation at UAI 2024.
• I compiled a set of slides on causal discovery, part of the teaching material of the Computational Epidemiology course in the MSc Health Data Analytics and Machine Learning at Imperial College London.
• A paper on causal discovery with unobserved confounding was accepted for presentation at AAAI 2024.
• A paper on bandits and transportability was accepted for presentation at NeurIPS 2023.
• I am part of the organizing team of the NeurIPS-22 workshop A Causal View on Dynamical Systems. Consider attending and submitting your work.
• A paper of ours on counterfactual estimation with time series data was accepted for presentation at ICML 2022.
• A paper of ours on graphical modelling with stochastic processes was accepted for presentation at ICLR 2022.
• I am part of the logistics committee of the NeurIPS-21 Workshop "Causal Inference & Machine Learning: Why now?" (WHY-21). Consider attending, see more details here.
• The recording of my presentation on synthetic controls at the Rice ECE Speaker Series Seminar is available.
• A paper of ours on causally-inspired imputation was accepted for presentation at NeurIPS 2021.
• Two recent papers of ours on hypothesis testing with sets and with selection bias were accepted for presentation at UAI 2021.
• One paper of ours on counterfactual estimation using synthetic controls was accepted for presentation at ICML 2021.