Benoît Pélissié du Rausas

Feasibility and Development of a Demand Forecast Model for Raw Material Needs • Designed, developed, and evaluated a multi-layer time series forecasting model using TFT, DeepAR, XGBoost, CatBoost, and Prophet, improving demand prediction accuracy and reliability. • Built a forecasting pipeline with automated data processing, model training, testing, updating, and performance tracking to ensure scalability and maintainability. • Performed EDA, data cleaning, and feature engineering to ensure data quality and uncover demand patterns and seasonal trends. • Developed interactive Streamlit dashboards to visualize data insights. • Conducted a feasibility study demonstrating the scalability and automation potential of the forecasting solution across product lines.

• Conducted customer segmentation using advanced analytics (RFM, clustering) to optimize marketing strategies. • Developed classification models using CNNs and NLP (BERT) for product categorization and automated quality control. • Built predictive models (XGBoost, Random Forest) for energy use and credit scoring, deploying dashboards (Streamlit) for real-time insights. • Implemented end-to-end MLOps solutions for API deployment and model monitoring using MLflow and Docker. • Leveraged big data tools (PySpark, AWS) for cloud-based processing and compliance management. • Applied computer vision techniques (YOLO, SegFormer) for automated defect detection and navigation solutions.

• Collaborated with stakeholders to determine relevant metrics and graphs to produce, identify data needs and develop strategies to answer business questions effectively. • Produced and automatized SQL queries to extract and present said data, focusing on customer lifetime value (LTV), average order value (AOV), and purchasing behaviors. • Created dashboards in Mixpanel to visualize key metrics and trends.

Business Intelligence & Marketing intern.