2011-2016 — The Academia
I was originally trained as a neuroscientist. I did my PhD at the CIRB of Collège de France with Alexander Fleischmann (now in Brown), studying olfactory processing in mice (see here for our three minutes of fame).
This is where I learned how to address problems scientifically, drive R&D projects, anticipate caveats and mitigate risks, monitor progress and communicate results.
We published two papers on odor coding in the mammalian olfactory system.
2016-2017 — Around the world
I took off with my wife and visited 17 countries in 11 months using plane, train, boat, hitchhiking, and on foot.
We experienced a great variety of environments and culture, met and interacted with people from all over the planet.We logged our impressions, recommendations, and photos in a blog (in French).
2017 - Almost launched a start-up
I kicked off an edutech start-up with my former colleage Thomas Deneux.
This is where I learned how to pitch a project to prospects and partners, and use their feedback to shape product ideation.
Learning Robots is using Deep Neural Networks to build fun and educative toys.
2018 onwards — Applied Data science and Machine learning in the industry
I then started applying advanced statistical methods in the industry as a data scientist consultant in the TP Lab (ex-Tunnel lab) of Bouygues Travaux Publics.
I used big data analysis and advanced statistical methods to improve civil work project efficiency.
This is where I learned that no data or machine learning project can ever begin without a deep understanding of the underlying business context.
In 2020, I joined Rubix (now Ellona), at first in charge of customer data operations and analysis, and later led the R&D projects to improve the company's anomaly detection and odor recognition algorithms.
This is where I learned the difficult art of delivering fancy data and machine learning products that really address customer needs.
I joined Zefir Pricing Accuracy team at the end of 2022, shipping data and ML features to improve recommender systems used by the backoffice for real estate valuation.
This is where I learned that uncompromising code quality and data/ML ops standards are key to reliable data and machine learning services.