Research Data Scientist - Internship - (F/H)

Location 92-HAUTS-DE-SEINE, France
Experience level Internships
Job details sector Big Data Science
Apply before Date not available


Performances of machine learning models are usually evaluated with aggregated metrics like the accuracy score. These metrics are not sufficient to assess a model and we may want to answer more complex questions such as:

  • What has been learned by the model? Why a particular prediction is made?
  • Where the model is right or wrong?
  • What can be done to change a prediction?
  • Is the model robust? Is it causal? Is it fair?
  • Etc.

Many machine learning models are black-boxes: answers to these questions are not simple to obtain and require the development of specific techniques, which is the objective of the machine learning interpretability field.

Interpretability of machine learning models is an important topic to develop the use of machine learning in organizations like AXA. The lack of model’s transparency is an issue when it comes to gaining business people’s trust in the predictions made by a model, to extract knowledge and to audit the quality of the models to improve them. It is also an issue from a legal and ethical perspective, in particular with the requirements of the new GDPR’s right to explanation (new European Union regulation starting in 2018).

The R&D team of the Data Innovation Lab at AXA conducts research to provide solutions and new methods to make machine learning interpretable.



The R&D team proposes an internship in machine learning interpretability whose objectives are:To contribute to the development of a library gathering state-of-the-art approaches to perform machine learning interpretability. This library will be used by data scientists of AXA and for the research conducted in the R&D team.


Depending of the will of the intern and the time available, one or two of the following:

 Research track: in line with the research performed in the R&D team, contribution to an effort of organization of the interpretability field, in particular by benchmarking the different approaches.

Applied track: application of the developed library to data science projects at AXA.


The candidate will benefit from a dynamic scientific environment: Data Innovation Lab with highly skilled Data Scientists, R&D team with PhDs in AI, a PhD student, Joint Research Initiative on ML interpretability with UPMC-LIP6 computer science lab…


  • Master 2 in computer science, mathematics or equivalent (engineering schools…).
  • Strong background in computer science, machine learning and mathematics.
  • Interest for academic research, in particular: read and understand an academic paper to translate it into code, design and conduct experimentations.
  • Fluent in Python (or ability to learn it quickly).
  • Fluent in English or French.

Would you like to wake up every day driven and inspired by our noble mission and to work together as one global team to empower people to live a better life? Here at AXA we strive to lead the transformation of our industry. We are looking for talented individuals who come from varied backgrounds, think differently and want to be part of this exciting transformation by challenging the status quo so we can push AXA - a leading global brand and one of the most innovative companies in our industry - onto even greater things.

In a fast-evolving world and with a presence in 64 countries, our 165,000 employees and exclusive distributors anticipate change to offer services and solutions tailored to the current and future needs of our 107 million customers.


The headquarters of the AXA Group, based in Paris 8th, brings together the Group's corporate activities. It coordinates the various entities with the Group's strategy, and is responsible for managing international projects. The headquarters has approximately 800 employees and is distinguished by its strong international culture (39 nationalities).