Max Cairney-Leeming

I am a PhD Student at the Institute of Science and Technology Austria (ISTA), near-ish Vienna, under the supervision of Christoph Lampert. I work on Differential Privacy for (deep) Machine Learning, and am interested in the intersection between DP and other concerns about machine learning, e.g. fairness, and also in the tradeoff between privacy and achieving good utility, in terms of where different privacy definitions/frameworks might find a different balance to what we have already. I am also a member of the ELLIS PhD Program, and so co-supervised by Amartya Sanyal at the University of Copenhagen.

Projects Notes

Preprint available: Private Worst-Group Optimization

Our recent work on “Adaptive Sampling for Private Worst-Case Group Optimization” is now available on the arXiv at arXiv:2602.10820. This work was jointly supervised by Amartya Sanyal and Christoph Lampert, and was began during my visit to Copenhagen.

Visit to the University of Copenhagen

I am visiting Amartya Sanyal’s group at the Datalogisk Institute at UCPH for the next 3 months, as part of my ELLIS exchange.

PPML Workshop @ EurIPS

I presented a poster on a preliminary version of my latest work on private worst-group optimization (now available at arXiv:2602.10820) at the PPML workshop, organized by several of my group and close collaborators.

Qualifying Exam

I presented my PhD project proposal to my committee: Christoph Lampert, Marco Mondelli, Amartya Sanyal and Florian Praetorius (chair).

Affiliated with the Lampert group

I have officially joined the group of Prof. Christoph Lampert at ISTA.

Starting my PhD

I have started my PhD at ISTA! I’m looking forward to working with 3 different professors over the next year, to broaden my ML horizons, and conversely, to hone in on a topic and affiliate with one of them for the rest of my time as ISTA.

Graduation

I graduated from Oxford! I recieved a First (years 1-3; Bachelor’s) and a Distinction (4th year; Master’s) in my integrated MMMathCompSci degree in Mathematics and Computer Science.