Jean Vassalo
About
Jean Vassalo is from Greater Paris Metropolitan Region. Jean works in the following industries: "IT Services and IT Consulting". Jean is currently Machine Learning / Python developer at Groupama, located in Suresnes, Île-de-France, France. In Jean's previous role as a Master Thesis in Data Science at Sahar (https://sahar.fr/), Jean worked in Ville de Paris, Île-de-France, France until Aug 2021. Prior to joining Sahar (https://sahar.fr/), Jean was a Research Intern in Topological Algebra at IRIF (Institut de Recherche en Informatique Fondamentale) and held the position of Research Intern in Topological Algebra at Paris, France. Prior to that, Jean was a Data Science Internship at ClaraVista, based in Paris, France from Jun 2017 to Aug 2017. Jean started working as French military police at Gendarmerie Nationale in Haute-Marne, Grand Est, France in Sep 2015.
Jean Vassalo can be found on Finalscout.com, where members can access Jean Vassalo's email for free. Finalscout is a professional database with more than 500 million business professional profiles and 200 million company profiles.
Jean Vassalo's current jobs
In order to make the administrative process more efficient, Groupama has developed an algorithm that uses deep learning character recognition (OCR) and other computer vision techniques to automatically read administrative documents sent by customers (such as ID card and driver's licenses). In this context : - I understood and rewrote in Python an existing C++ algorithm for document recognition - I conceived a new version of this algorithm, more robust to noise and outliers, which I've then tested - I coded a Python package that synthesize ultra realistic pictures of documents using multiples effects such as (blur, shadows, contrast, jpeg compression, change of perspective etc...). These pictures were synthesized according to different template so that it can precisely imitate classic documents such as ID card of passport. - I adapted the previous code so that it could also synthesize images representing text written on landscapes pictures. To do so, I used a Python script that automatically saved thousands of pictures on the internet reading the htlm code of some web pages (web scraping) - I made this package more efficient then I used it to generate a dataset to train a deep learning character recognition model (OCR) with PyTorch
Jean Vassalo's past jobs
Unsupervised learning for marketing purposes on Twitter data. In order to detect communities of users on a large Twitter graph : - I have computed a dataset of shortest paths on a massive Twitter graph using a multiprocessed version of Dijkstra's algorithm. - I have then trained and fine-tuned different unsupervised machine learning algorithms (k-means, gaussian mixture) on this dataset. - Finally I've used the trained model to detect communities of users on Twitter, sharing the same interests, for marketing purposes.
Topological and algebraic tools applied to Artificial Intelligence. I gained knowledge in topology. Then with Doctor Yves Guiraud, I found a different proof to an existing theorem which I used to try to find a new machine learning algorithm in the field of topological data analysis.
Machine Learning applied to finance, in collaboration with Pharo Management. I've implemented a dataset of stock market values in Scala (filling missing values and computing new variables). To do so, I had to optimize the speed of SQL requests. Then I trained machine learning models with Tensorflow to predict the USD/JPY exchange rate.
After three months of training as an officer in the National Gendarmerie Academy (EOGN), I joined Haute Marne's Groupement as a Commander Assistant for leadership training, and to help to the Groupement's management.