Mamar O.

- Training of an ASGD, Passive Aggressive, Naive Bayes, Random Forest & Ada Boost model and experimentation to determine which is best for a document classification task based on relevant keywords, according to the F1-score metric. Use of the TF-IDF method for data normalization and Grid Search for hyperparameter optimization.

- Retrieval of 30,000 image examples of various data structures found in a document (formula, table, form, signatures, etc.) to train an internal company model for this image classification task.

- Development of a JAVA program consisting of an ETL. Raw data was provided via batch and updated every evening in the form of three different CSV files. The program mainly consisted of retrieving these files before performing multiple transformations such as date formatting, deletion of certain data, addition of a column whose value for each row depended on the value of other columns, and more. Each file had a specific processing and the program was divided into three different scripts: one for each file.

- Use of web technologies via Java to create a page for an internal company tool. This tool consisted of displaying data contained in the database. SQL query work involving joins and groupings was necessary.

- Development of an HTML & PHP website and implementation of a database so that a company technician can record interventions performed at clients' sites via a form. It will be possible to view these interventions (with a filtering system), modify or delete them, as well as send them directly in PDF format by email to the client with a pre-filled signature.

- Showcase website for a merchant's maintenance services.