Saifuddin Syed

Saifuddin Syed

Department of Statistics

University of Oxford


I am a postdoctoral research assistant in computational statistics and machine learning at the University of Oxford supervised by Arnaud Doucet and funded by the CoSInES project. I recently have joined the algorithms and inference working group in Next Generation Event Horizon Telescope (ngEHT) collaboration to help improve the algorithms used to model and image supermassive black holes. Before this, I completed a PhD in Statistics with Alexandre Bouchard-Côté at the University of British Columbia.

My research aims to design algorithms for solving Bayesian inference problems in the applied sciences. I am interested in developing general-purpose algorithms that are mathematically grounded and can efficiently scale to modern computing resources. My work is interdisciplinary and lies at the interface between probability theory, statistical physics, machine learning, and differential geometry.

During my PhD I studied a non-reversible variant of a popular algorithm in computational statistics and physics called parallel tempering. We showed that the non-reversible parallel tempering (NRPT) dominates the traditionally used reversible counterpart, can be efficiently tuned and can scale to GPUs. See the applications section for recent examples of projects using NRPT at scale and the publications section to learn more about NRPT.

  • Parallel tempering
  • Monte Carlo methods
  • Bayesian inference
  • Statistical physics
  • Stochastic analysis
  • Information geometry
  • PhD in Statistics, 2022

    University of British Columbia

  • MSc in Mathematics, 2016

    University of British Columbia

  • BMath in Pure & Applied Mathematics, 2014

    University of Waterloo

Notable Applications

Here are some recent examples of large-scale projects that used our NRPT algorithm as the backbone of their inference engine.

Generating first image of Sagittarius A*
The Event Horizon Telescope collaboration used NRPT to image Sagittarius A*, the supermassive black hole at the center of the Milky Way.
Modeling evolution of single-cell cancer genomes
NRPT was used to study the fitness of cancer populations by modelling the evolution of single-cell cancer genomes.
Modelling plasma dynamics in nuclear fusion reactor
TAE technologies in collaboration with Google research and TensorFlow probability used NRPT to infer plasma dynamics inside in the world largest field-reversed configuration (FCR) nuclear fusion reactor.
Discovering magnetic polarization in M87 supermassive black hole
NRPT was used by the Event Horizon Telescope collaboration to discover magnetic polarization in M87’s supermassive black hole.


(2022). Local Exchangeability. To appear in Bernoulli (Accepted).

PDF Cite Arxiv

(2022). Parallel tempering with a variational reference. To appear in the Conference on Neural Information Processing Systems (Accepted).

PDF Cite Poster Arxiv

(2021). Non-reversible parallel tempering: a scalable highly parallel MCMC scheme. Journal of the Royal Statistical Society (Series B).

PDF Cite Poster Video Source Document Arxiv

(2021). Parallel tempering on optimized paths. In the International Conference on Machine Learning.

PDF Cite Poster Slides Video Source Document Arxiv

Recent talks

Harvard University Black Hole Institute Seminar
ISBA 2022 World Meeting
Université de Montréal Statlab Seminar
CoSInES Blue Sky Kitchen 2022
University of Oxford OxCSML Seminar


  • +44 7467 304999
  • 24-29 St Giles, Oxford, CA OX1 3LB
  • Office 1.18