Kais Laribi

Projects related to data science and ML engineering: AXA GO: As an ML Engineer, I support teams in establishing best practices for ML and GenAI projects. I also worked on analyzing LLM risks and implementing guardrail solutions for customer-facing applications to enhance trust in AI systems. Sanofi Accelerator: Engaged in projects related to Marketing Mix Modelling. These projects encompass the creation and ongoing management of Bayesian Marketing Mix Models (MMM), gaining insights into historical performance through the generation & analysis of response curves, and enhancing the allocation of marketing budgets for optimization purposes. Brut media: Designed a robust architecture leveraging Google Cloud Platform (GCP) to streamline the deployment of machine learning models and seamlessly integrate them into the BigQuery environment. The primary objective of this project was to automate the deployment workflow for data scientists, enabling them to serve models effortlessly. Additionally, the architecture aimed to provide analysts with easy access to these machine learning models through SQL queries within DBT / BigQuery. - Developed comprehensive training sessions for the data team, focusing on utilizing data-driven approaches to create effective marketing use cases. The training encompassed a wide range of topics, including various use case ideas and methodologies required to tackle them successfully.

- Built an audience segmentation methodology for our publishers (websites with +30M visitors/month). - Set best practices and MLops tools to deliver data science/ML projects. - Synchronise with the business, analytics, engineering teams to translate needs into technical solutions.

Working in the research and innovation department on traveler behavior modelling when choosing airline tickets using statistics and machine learning models. Keywords: choice modelling, statistics, data analysis, market segmentation, machine learning

The project aims to implement an end-to-end text detection and recognition system using deep learning models from literature. Basically, we dealt with the detection problem using object detection models ( faster RCNN, SSD...) trained on text images datasets. For the recognition task, we implemented a sequence to sequence OCR model that uses convolution and LSTM layers to predict words corresponding to cropped images. Keywords: Computer vision, deep learning, transfer learning, tensorflow, OCR

Customer lifetime value prediction using Deep learning techniques: We used deep learning, particularly recurrent neural networks to understand customer behaviour in the retail sector. To achieve this, We implemented different RNN models to predict customers' lifetime value using the RFM parameters ( well known parameters in marketing ). The project is a part of a kaggle competition : https://www.kaggle.com/c/acquire-valued-shoppers-challenge

Programming a path remembering robot. -Implementing ROS/C++ nodes to store a trajectory -Adding a control method to guarantee following stored trajectories Keywords: C++, Ros

Workman internship carried out in the workshop of TUNITERM: a company specialized in manufacturing thermal sensors and thermistors.