Job Description

Are you a computer scientist with an interest in stochastic modelling within a causal inference framework? Are you interested in applying stochastic dynamical techniques to data science?  Are you a hands-on software developer with experience in C/C++, Python, R and CUDA?

This project will investigate the improvement of standard prediction methods in statistics. Traditional prediction methods often ignore the existence of subpopulations in the data and so-called ‘enigmatic variation’ – the idea that otherwise identical individuals can follow considerably different trajectories based on noise they experienced at some point. You will investigate the improvement of prediction methods within a casual inference framework that allows the inclusion of context-dependent information and that takes account of how the data were generated.

We hypothesize that the inclusion of temporal dynamics and explicit modelling of inhomogeneities in the population will lead to improved prediction. The aim of the project is to deliver a working prototype of a software implementation of the improved methods and to investigate its performance. You will adapt an existing simulator, based on a combination of C/C++, Python and CUDA for use in a causal inference framework. You will use the simulator to investigate the performance of so-called Directed Acyclic Graphs (DAGs) augmented with so-called population density methods and produce a written evaluation of the results. The project is sponsored by the Alan Turing Institute on behalf of a third-party funder.

What we offer in return

And much more!  

To explore the post further or for any queries you may have, please contact: 

Dr Marc de Kamps, Associate Professor

Tel: +44 (0)113 343 5322 or email: [email protected]