I am an Assistant Professor of Statistics at Harvard University. At the interface of Statistics and Machine Learning, I develop scalable and principled methods for inference in complex scientific models. My research proposes new ways of parallelizing Monte Carlo methods, of comparing statistical models based on predictive performance, of retrieving latent processes given observations, of performing inference in non-Markovian and nonlinear models, and of learning with purely generative models with intractable likelihoods. These contributions aim at helping practitioners of data analysis in Science and Industry, by providing efficient computational tools to calibrate and compare complex models, and ultimately, to facilitate decision-making in the face of uncertainty.

More specifically, I look into:
  • Monte Carlo algorithms with parallel computing in mind;
  • nonlinear dynamical systems, state space models;
  • sequential inference (i.e. updating inference as data points arrive);
  • purely generative models and inference with intractable likelihoods using ideas from optimal transport;
  • Bayesian inference in models made of modules, model misspecification.
Upcoming talks:
  • University of Toronto, November 16, 2017.
  • NIPS workshop on Optimal Transport & Machine Learning, December 9, 2017.
  • O'Bayes '17 at the University of Texas in Austin, December 11-13, 2017.

I am not taking any interns over the summer. If you're interested in a PhD, look for info on the Department's website.

Email: pjacob at fas.harvard.edu.
Phone: (617) 496-9259
Office: Science Center 712, 1 Oxford Street, Cambridge, MA 02138