Jany Lahlouh

• - Retail: Scaling up daily sales forecasting and model monitoring. Exploring the capabilities of LLM for industrial forecasting. Over 180,000 series are automatically forecasted every day.

- Retail: Won a hackathon against 4 firms, leading to a partnership, starting with a multi-product and multi-site sales forecasting project. Implementation and structuring of the data science/ML project and workflow for a full team. Modeling involving in-depth data exploration (e.g., cannibalization), advanced feature engineering, various models (GBDT, RNN, TCN, CHRONOS/LoRa), and deployment for daily predictions. Project carried out on Databricks. Support for less experienced data scientists.

- Internal: Ideation and implementation of an internal Python ML library adapted to recurrent challenges faced by teams.

- TV: Development of a dynamic pricing formula (GRP).

- Retail: Price elasticity study in a mass-market retail context aiming to prepare for a dynamic pricing project.

- Medical: 3D radio image segmentation (dicom) using deep learning, implementation of a UNET-R challenged by totalsegmentator. Docker deployment.

- Agriculture: Prediction of heat and calving periods for cows from multivariate time series, involving advanced feature engineering, gradient boosting, and recurrent deep learning models, followed by AZURE deployment.

Within a Deep Learning R&D team, I had the opportunity to work on a computer vision problem: Detection, Recognition, and Identification using self-supervised learning and active learning.