Nicolas Mormiche

• Pedagogical support and coaching

• Teaching assistance and problem-solving

• Tutoring and project progress monitoring

• Topics covered with learners:

SQL · Python · Pandas · Matplotlib · Seaborn · Plotly · Scikit-Learn · MLflow · Docker

Machine Learning · Deep Learning · Regression · Classification · Time Series

• E-commerce product classification and type code prediction.

• Import, exploration, and data visualization of data with 84,916 products.

• Cleaning and transformation with 'prdtypecode' as the target variable and 'descriptif' for text and 'filepath' for the associated image as explanatory variables.

• Text pre-processing (NLP): removal of excessive spaces, HTML tags, special characters, and stopwords, then language detection.

• Image pre-processing (Computer Vision): format conversion, color adjustment, resizing, filter application, and data augmentation.

• Transformation of textual data into numerical data using tokenization, lemmatization, data augmentation, and then vectorization to improve the performance of future models and avoid overfitting as much as possible.

• Training of advanced machine learning models (Naive Bayes, KNN, SVM, and DecisionTree) and deep learning models (Sequential, CNN, ResNet50, and BERT).

• Model analysis with metrics: accuracy, f1-score, and loss function.

• Model interpretation with LIME to explain predictions.

• Improvement of scores by 5% with 0.81 for text and 0.46 for images.

• Development of back-end features for nearly 50 projects including showcase sites, e-commerce, CRMs, and CMSs in various sectors (banking, insurance, pharmaceuticals, telecoms, gaming, luxury).

• Support manager responsible for handling maintenance and evolution requests.

• Lead developer for a team of 4 on a nationwide B2B store.