flora bajolet geodata
Credit scoring model, dashboard and API
Use bank and credit data of "Home Credit" company to build a model quantifying the clients default risk. This is a supervised binary classification problem. A dashboard and API are deployed to help bank advisors in their decisions.
Deploy a scalable model on the cloud
Here we seek to develop a computer vision model able to classify fruits and vegetable photos. The model must be able to run on thousands of images which requires a spark/hadoop set-up deployed on a cloud plateform. For now, only a POC including the loading of images, preprocessing and dimensions reduction is needed.
NLP and computer vision for e-commerce goods
This project aims at classifying items from a title/description + one image of the product. I tested several NLP and computer vision approaches, both with unsupervised, supervised and transfer learning approaches.
Clients clustering to drive marketing actions
A brazilian e-commerce company wants a clients database analysis to provide insights for their marketing team. I use clients personnal data in RFM and relational clustering to highlight their shopping behaviours. On this project, I focused on getting a clear view of the clients pool rather than technical aspects of modeling.