Panos Tsimpos

MIT ORC & MIT LIDS

I am a PhD student in the Operations Research Center (ORC) and the Laboratory for Information and Decision Systems (LIDS) at MIT advised by Professor Youssef Marzouk. My research focuses on probabilistic machine learning, with an emphasis on generative modelling, sampling and dynamic measure transport. I work on both theoretical foundations as well as novel algorithmic paradigms, drawing on tools from stochastic analysis, modern statistics, and mathematical physics. I am currently especially interested in the principled design of probabilistic generative models.

Before MIT, I studied Mathematics and Physics at Columbia University.

You can reach me at ptsimpos@mit.edu.

news

May 11, 2026	New preprint out: One Operator for Many Densities: Amortized Approximation of Conditioning by Neural Operators, proposing an operator learning framework for conditional inference. This is joint work with Edoardo Calvello, Ayoub Belhadji and Nicholas H. Nelsen. Find the pre-print here.
Apr 16, 2026	New preprint out: One-Shot Generative Flows: Existence and Obstructions, studying a central question in the design of probabilistic generative models. This is joint work with Daniel Sharp and Youssef Marzouk. Find the pre-print here.
Mar 22, 2026	Gave an invited talk titled Design of dynamic measure transport: a study of one-shot sampling at the SIAM Conference on Uncertainty Quantification (UQ26) in the mini-symposium Measure flows for inverse problems and machine learning.
Jun 30, 2025	Gave a talk titled Optimal Scheduling of Dynamic Transport at the Conference on Learning Theory (COLT 2025).
Nov 24, 2024	Gave a chalk talk titled Optimal Schedules in Affine Dynamic Transport at the Oberwolfach Seminar: Control and Machine Learning.