Léa Cochet

Extraction of information from contract termination documents (NLP) : - Data clustering to detect different formats : k-means method - Data preprocessing using BERT vectorization - Training neural network models CNN and LSTM - Monitoring the model in production Table detection on insurance history statements (computer vision) : - Application of the Table Transformer (TATR) deep learning model, based on object detection Technical environment: Python (pandas – scikit learn – tensorflow – transformers – pytorch – regex - kedro - streamlit), Git

- Extraction of anomalies from satellite health reports and severity classification (NLP) - Data preprocessing by comparing TF IDF and Word2Vec vectorization - Modeling through semi-supervised learning : random forest, gradient boosting, SVM and neural networks Technical environment: Python (pandas - scikit learn - tabula - mysql - nltk - regex), Git

- Modeling the number and cost of claims for the implementation of a multi-risk professional insurance product (GLM models) - Drafting the insurance product note including target identification and market study Technical environment: R (ggplot2 – caret – bestglm), Excel

- Clustering of bacterial resistance gene sequences - Hierarchical classification, PCA (Principal Component Analysis) Technical environment: R, Excel

- Analysis of promotions from a large french retailer - Formatting databases Technical Environment: Excel