Time for a change

December of last year I left my position as a technical consultant at Quantmetry, to join the INRIA center in Rennes, Britany. Being a consultant was a great experience : at the time I just came out of school (finally) and wanted to confront my academic knowledge to real-world use cases, which I did, in different companies, in different sectors. From this point of view consulting is good to start a carrer with when one wants to understand how big industries work, and Quantmetry was definitely a great place for that.

This time as a consultant was also the moment I got introduced to research, which included reading papers along with writing blog posts and white papers. But as consultants we often have to fulfill plenty of different tasks, responding to different needs for high demanding clients, leaving research projects sometimes a bit left aside, sadly.

I had an amazing time at Quantmetry, I just felt like it was the time for me to focus on more long-term, wide-scale projects.

How I pictured the INRIA

It had already been a few years since I’d heard about the INRIA, back in my first internship when I first discovered scikit-learn, and the brand new world of machine learning. I quikcly started to browse the internet and read many many resources on scikit-learn, and started to follow its main protagonists, especially Olivier Grisel and Gaël Varoquaux.

During my second internship, in 2016, I discovered dask, and kept following the project eversince. The project grew astonishingly quick, and I witnessed how people at Anaconda interacted with the already well-established scikit-learn comunity. A consultant at Quantmetry I continued to follow both projects, and in 2019 I attended the scikit-learn consortium.

For a long time, that was it.

Moving to Rennes

I joined the center of Rennes in early December of 2019. I was assigned 2 exciting projects.

Pattern mining

The fist one deals with pattern mining, and aims at bringing those underestimated techniques of data-mining to the forefront via a dedicated library, while making interactions with sklearn as easy as possible so that users can mix approaches.

It is a really exciting project for many reasons

  • First of all we start from scratch. Nothing really exists to serve this purpose in the python ecosystem. Even choosing the datastructures to use is a topic in itself, because there is no concensus as there is in machine learning.

  • Secondly, we draw on 20 years of research (the first algorithms were developped in the late 90s), so I need to keep a high level view to identify how we could serve the common interest in the best way.

  • Last but not least, from a scientific point of view this project is amazing. As pattern mining is a form of symbolic Artificial Intelligence, I remain curious to see to what extent the excitment around such methods can go, either being for the methods themselves or for using them with more trendy connexionist methods, namely ML and DL algorithms. Taking part of this project position me halfway between the two main types of AI, and make me a more accomplished and cultivated scientist.

Neuro imaging

The second project I am working on is about provenance description in the field of neuro-imaging. This is completly different as this project is at the crossroads of many mature teams, composed of senior researchers working all around the globe.

I feel a bit like a noob at the moment, but this is a great occasion for me to learn about many things : what tools are used in neuro-imaging, workflow standards, how the teams are working, funding systems, and so on.

Long time no see

Research takes time.

Event if I work as an engineer - not as a researcher - I am amazed how much time and energy have been allocated to my onboarding process. Team members regularly ensure everyone is on the same page, and I have been given plenty of time to read papers and other materials online. I had never been reading so many papers !!

Even as an engineer I need to stay committed into research processes, which makes some general rules appliable to PhD students and senior researchers also appliable to me. On a daily basis work needs to be defined quite seriously, so I don’t branch off completly from my original subject and end up consuming time for nothing. But all of these side effects are offset by frequent communication between members.

My first few months have been quite surprising, but I do not regret my choice to have entered such an arena and feed my curiousity. I find my work interesting, both on a professional and a personnal level. I also find people very humble, despite academic success or seniority.