Data Scientist2016 - PRESENTCorelyo
Technologies: Python, Sklearn, Keras, NLP, Data Analysis
- Implemented a machine learning algorithm for document classification. The full pipeline included OCR (optical character recognition) to extract features from pdf and scans, feature engineering and cleaning for bad quality scans, and an occurrence-based classification model.
- Worked on various analysis and business intelligence (BI) projects. This included assessing the impact of ads on users of an eShop, creating a complete set of KPI/charts for a CRM editor, and building dashboards for HR indicators.
- Mentored students for openclassrooms and hold weekly student appointments to guide them through their data science journeys.
- Designed from scratch a recommendation engine for a B2B eCommerce website selling hardware to contractors. Tested several architectures and models, and set up an API to serve recommendations when needed. Also implemented a docker container for re-training the model on a regular basis.
- Developed an unsupervised image classification algorithm based on image similarities, applied on social network images. The objective was to evaluate the social engagement generated by each type of image, to optimize the social presence of my client.
- Worked on an exploratory data analysis on tennis data, aiming to discover the importance of the first serve percentage in a game, and more broadly what can give a player an edge over its opponent. The deliverable was a report with mostly visualisations and interpretation of my results.
- Implemented a clustering API for SEO keywords, based on the SERP results of these keywords. The goal was to make connections between keywords that are syntactically very different but of same meaning.
- Designed, for a growth hacking agency, an algorithm to find common traits between prospects and clients and used this to prioritize prospects and find out which product they were most likely to be interested in.
Research Engineer2014 - 2016LML
- Developed a Python module for testing materials which enabled PhD students to quickly design custom setups with synchronized sensors, camera, and actuators.
- Conducted some tests on my own and extracted useful data from the measures.