Saad Kabbaj

– Developed of a conversational AI chatbot enabling users to interact with their dashboards and obtain personalized insights. Implementation of MCP protocol as a security layer ensuring per-user data isolation and protection against SQL injections, even in cases of LLM hallucinations. Technologies: Azure AI Foundry, Databricks, Microsoft Fabric. – Designed and deployed a machine learning algorithm to predict delivery times for orders on the Marjane Mall platform, enabling improved customer experience and logistics optimization. – Created an automated recommendation matrix to suggest alternatives for out-of-stock products on Marjane e-commerce, leveraging natural language processing (NLP) techniques to analyze product characteristics and similarities.

– Designed an IDP (Intelligent Document Processing) application to make back-office processing tasks more efficient, by implementing advanced data extraction techniques, using state-of-the-art vision language models, and deep learning models. Focused on extracting complex elements like formatted content, graphical elements (stamps, signatures), and handling noisy images and some handwritten text. – Designed and executed data-driven strategies to assess and predict consumer credit appetite, enabling tailored credit offerings and optimizing risk management. – Engineered a sophisticated recommender system to match customers with the most suitable financial products, improving customer satisfaction and product uptake.

– Designed and implemented an insurance fraud detection model, improving the overall process efficiency.

– Contributed to the enhancement of the financial risk management process through the recalibration, development and maintenance of the financial risk models. – Reported key risk indicators of the bank for the different committees. – Monitored that the indicators are within the predetermined limits according to the bank's risk appetite. – Improved data quality & performance by analyzing, verifying and modifying SAS & Python scripts – Developed and implemented methods to make machine learning models interpretable for non-technical stakeholders.

– Designed and built a predictive model of fertilizer consumption in China over 10 years as an intern for the OCP Group. – Used time series analysis techniques, including Fb Prophet, to analyze historical data and forecast future trends in fertilizer consumption. – Employed linear regression and ensemble learning techniques to improve the accuracy and robustness of the model. – Conducted feature engineering and selection to identify the most relevant variables for predicting fertilizer consumption in China. – Worked closely with the strategy team at OCP Group to understand their business needs and ensure the model aligned with their goals. – Presented the model and its findings to the OCP Group's senior management team and provided recommendations for future actions. – Conducted sensitivity analysis to identify the most critical assumptions and inputs that affected the model's performance.

– Co-authored a research paper on Vehicular ad hoc networks after doing numeric simulations of 4 proposed routing protocols in industrial environments.