👋 Hello!

In April 2022, I moved to London, UK and joined Integrum ESG as the Machine Learning Lead. Currently, I am mainly focusing on building NLP models using methods derived from Signal Processing theory. My workdays are also spent thinking about ML scalability as I am fortunate enough to work with (very) large data sets.

Prior to joining, I worked on the connections between Causal Inference, Machine Learning and Data Science at Harvard, where I was a Research Fellow in Prof. Tyler VanderWeele’s group, and at the University of Washington (Seattle, WA) in Prof. Tim Althoff’s Behavioral Data Science Lab.

I graduated from ENSAE, Institut Polytechnique de Paris with a MS in Mathematical Statistics and from ESSEC Business School with a MS in Economics and Management. During my studies, I was a data scientist intern at Quantmetry.

👀 Research interests

I am very broadly interested in Causal Inference, Statistical Methodology for Complex High-Dimensional Data and Explainable, Robust and Fair ML.

Recently, I enjoyed using NLP, matching techniques and sensitivity analysis in the context of social media and mental health - here are introductory slides from a brief talk I gave at UW in September 2020. In the past, I also had fun working on unsupervised reinforcement learning, generative adversarial networks, parallel computing, auction theory and times series forecasting.

In my free time, when I am not cooking, I like reading and (occasionally) writing about sociology - Essay 1 and Essay 2 are two recent attempts to write about the behavioral determinants of health and the politics of health inequality.

📝 Publications

• Deshayes, R., Murnaghan, T. (in press). Pioneering Transparent ESG Analysis on Public Markets: Leveraging Machine Learning to Transform ESG Data. ESG Club Conference. Portfolio Institutional.

• Captari, L. E., Sandage, S., Cowden, R.G., Deshayes, R., Choe., E., Crabtree, S., Davis, W., & Aten, J. (in press). Spiritual, religious, and existential processes of posttraumatic growth: Seeking, struggle, and transformation. Routledge International Handbook on Posttraumatic Growth. Routledge.

• Deshayes, R., Cowden, R.G., Shiba, K., VanderWeele, T. J., & Fancourt D. (in preparation). Associations of Underlying Motivation for Compliance with Public Health Measure During COVID-19 Lockdown with Subsequent Health and Well-being Among UK Adults. Expected December

• Cowden, R. G., Ho, S. Y., Deshayes, R., A., Cellini, J., Padgett, R. N., Stamatis, A., Morgan, G., & VanderWeele, T. J. (in preparation). Mapping the observational and experimental evidence on personal suffering and human flourishing: A scoping review. Expected December

🔍 Selected projects

Reinforcement Learning

• Unsupervised RL & Maximum Entropy RL: the DIAYN algorithm
  pdf code video

GAN

• Probability distribution learning and Generative Modeling - DCGAN & CycleGAN
  pdf code
• Optimal Transport GAN
  pdf code

Compressed Sensing

• Semidefinite Relaxations for Integer Programming - Max-Cut and Graph Coloring
  slides in French code

NLP

• Fake news detection with PyTorch
  pdf code

Bayesian methods

• Predictive view of Bayesian Clustering
  pdf code

Parallel computing

• Parallel Lasso in Cython
  pdf code