Oussama Zbir

As a Data Scientist & AI Engineer at BCP Technologies, I delivered end-to-end ML, GenAI, and big data solutions that supported key business decisions and improved customer analytics, credit targeting, and operational efficiency. Some projects delivered at BCP Technologies: • Agentic AI Trading Chatbot (Advanced RAG): Built a trading-focused agentic chatbot using LangGraph, routing agents, and agent-to-agent communication. Integrated document & web retrieval, voice interaction, conversation history, and a ChatGPT-like UI using Azure OpenAI, Qdrant, and Azure Document Intelligence. • Insurance Propensity Model: Developed a machine learning model to predict customer propensity for “Attamine Chaabi Hissab”, enabling targeted marketing and increasing adoption (Python, SQL, TensorFlow). • Customer Churn Prediction: Built and deployed a churn prediction model on a distributed Big Data pipeline (Spark, NiFi, Hive, Airflow), contributing to a reduction in customer attrition. • Branch-Level Credit Target Forecasting: Created time-series models (SARIMAX, LSTM, RNN) to forecast consumer credit targets per branch, integrating geography, seasonality, and macroeconomic factors.

Worked on a full end-to-end churn prediction project for individual clients at Banque Populaire: • Built and evaluated ML models for customer churn prediction (LR, RF, SVM, etc.). • Processed and analyzed large-scale customer data using Spark in a distributed big data architecture. • Applied techniques for handling imbalanced data (SMOTE, oversampling, undersampling). • Assessed model performance using AUC, precision, recall, and F1 score.

• Delivered end-to-end Machine Learning and Deep Learning solutions for global clients. • Built forecasting models, intelligent trading automations, and deep learning pipelines for computer vision tasks. • Designed scalable data workflows, ETL processes, and feature engineering for real-world production environments. • Developed and deployed applications using classical NLP, early transformer-based models, and modern deep learning frameworks. • Handled full project lifecycle: requirement analysis, data processing, model development, evaluation, deployment, and documentation.