Sofia Gilardini
DPhil — Statistics & Machine Learning
University of Oxford · StatML CDT
I'm a PhD candidate (DPhil) in Statistics and Machine Learning at the University of Oxford, part of the StatML CDT jointly run with Imperial College London. I'm supervised by Dr Kelly Zhang (Imperial), Professor Nick Hawes (Oxford) and Professor Yee Whye Teh (Oxford).
Before Oxford, I completed an MPhil in Machine Learning and Machine Intelligence at The University of Cambridge and a BEng in Electrical and Mechanical Engineering at The University of Edinburgh (First Class Honours), with an exchange year at UCLA. I care deeply about working on problems that are relevant in industry. I began my professional career working on robotic modelling and control at CMR Surgical, and more recently worked on robotics and automation within a computer vision algorithms team at Apple.
I also enjoy meeting people with similar ambitions, building at the frontier of what's possible. Alongside my DPhil, I am a Student Investment Partner at Creator Fund, a pan-European deep-tech VC fund that backs scientific founders emerging from university research.
I'm Italian-Argentine, and grew up in London and Madrid. I care deeply about people (generally) and love communicating with them in their native language. I speak Spanish and Italian natively, conversational French, and I am currently working on stringing together a coherent sentence in German.
Outside research, I learnt to row at Clare College, Cambridge. I love to hate rowing, and now row for the First Boat at Jesus College, Oxford. I also enjoy running and swimming.
Research
My first DPhil project explored generative modelling frameworks for forecasting from high-dimensional sequential data under irregular sampling. I am now working on bandit algorithms in nonstationary environments, where reward distributions shift over time. Prior to Oxford, I worked extensively on self-supervised learning methods for healthcare signals — including wearable IMU data and neural recordings — with a focus on learning useful representations from minimal labelled data.
Consulting
I occasionally take on technical consulting work for startups and early-stage companies — particularly at the crossroads of machine learning, physical systems, and engineering. If you're working on something interesting and think my background might be relevant, feel free to get in touch.