picture of meIn a nutshell I am machine learner who's main research interest is to uncover relevant and reliable data-driven metrics to evaluate solutions to a given problem. In particular I am very interested in online learning in general and in recommendation. Also as a former software engineering student (I hold a M. Sc. in Software Eng.), I generaly try to provide easy-to-use, reliable and extandable implementation of my research.

I just finished writting my PhD thesis at Lille 1 university. I studied under Philippe Preux and Jérémie Mary within the SequeL (Sequential Learning) research group (INRIA - LIFL).

In substance my thesis (which is available here) deals with the evaluation based on real data of adaptive recommender systems and classical recommender systems. At first my purpose was to derive adaptive recommender systems, better suited for dynamic environments. I even won an international challenge on the subject at the beginning of my thesis. Yet the problem of evaluating such systems turned out to be so interesting, so pervasive in my work and so important for the field that I dedicated all my efforts to it. Among other things I provided a thorough theoretical analysis of an existed method and contributed to the organization of a new international challenge, whose results revealed the extent of the shortcomings of that existing method. Based on that, I derived a new set of methods based on bootstrapping that both reduce the variance and the bias (which is rather uncanny) of the evaluations.

Besides my current scope of research, I really enjoy learning new things in various areas. For instance I recently learned how to use node.js, I'm currently taking an online course about PGM on coursera and I just read A brief history of time, or Big Bang/Quantum mechanics 101 as one may call it. I also really like passing along the little I know to others. I have had the exciting opportunity to teach software engineering as well as AI classes in Lille University and did so with great pleasure. I also sometimes tutor kids who struggle with their homeworks in my neighborhood.

Last but not least, I do enjoy nicely written code. Not just for the sake of it. To me poorly written code simply does not last long. And if it does, well... Hi trouble!