Megan
Stanley

I'm an AI researcher building foundation models and reasoning systems for scientific discovery. I currently lead on foundation model research within AI & Robotics at the Ellison Institute of Technology. Previously, at Microsoft Research, I was a first author on Aurora, a foundation model for the Earth system, built on a 3D Swin Transformer backbone with a Perceiver-based encoder, that set a new state of the art for weather forecasting and Earth-system prediction (published in Nature, 2025). I came to machine learning from physics, with a PhD from Cambridge, and have spent close to a decade doing AI research in industry.

I work across the full research lifecycle, framing the questions that matter, developing new methods and architectures, training and scaling the systems, validating them against data, and translating them into real-world impact. I lead research and mentor others, but like to stay hands-on: I still design models, write code, and run experiments.

My research centres on AI systems that build and refine internal models of complex domains, world models, and that improve through interaction with data, tools, simulation, and empirical feedback. Foundation models have dramatically expanded what AI can represent and reason about, but many scientific problems also demand systems that can form hypotheses, test them, and update against observation, experiment, and verification.

I'm particularly interested in reasoning systems, world models, foundation models, and scientific AI, and in architectures that couple learned representations with external tools, simulation, and verification. My goal is to build systems capable of robust reasoning, discovery, and adaptation, and to apply them to problems where they can create real scientific and real-world impact.