OMAR BOUKHERYS

• Telegram Agent: – The goal of this project was to develop an intelligent Telegram agent capable of answering customer questions in a relevant and automated manner. The agent can receive requests in text or voice message format, understand them through natural language processing, and then provide appropriate, fast, and consistent responses to improve customer experience and optimize support. – Develop Python code to extract and clean data from websites. – Build a vector database for this cleaned data. – Design an artificial intelligence system capable of interacting with users as an intelligent agent. – Create a user interface allowing end users to interact with the data. – Tools: AWS, Python, Qdrant, ElevenLabs, Groq, SQLITE, Langchain, Langgraph, Telegram, OpenAI API. • Serverless Recovery Dashboard: – This project aims to develop a serverless dashboard for monitoring and analyzing recovery operations, providing real-time insights and automated reporting capabilities to track key performance indicators and optimize recovery processes. – Design and implement serverless architecture for scalable data processing. – Develop automated data pipelines using Airflow for ETL operations. – Create interactive dashboards for recovery metrics visualization. – Integrate with Access DB for data storage and retrieval. – Implement data quality checks and validation processes. – Tools: Python, Airflow, Access DB. • Data Pipeline CTX Project: – This project focuses on building a robust data pipeline infrastructure to consolidate and process customer transaction data from multiple sources, enabling comprehensive analytics and reporting for business intelligence purposes. – Participate in meetings with client business analysts to better understand their needs. – Design and implement scalable ETL pipelines for data ingestion and transformation. – Develop Python scripts for data extraction, cleaning, and validation. – Orchestrate data workflows using Apache Airflow for automated scheduling. – Integrate multiple data sources into Access DB for centralized storage. – Tools: Python, Airflow, Access DB. • Customer Scoring for Auto CL Product: – This project aims to develop a scoring grid for customer evaluation, in close collaboration with the Risk team. The objective was to improve evaluation accuracy by refining existing criteria and implement a solution capable of providing real-time predictions. – Collaborate with the Risk team to understand current grid limitations and identify improvement areas. – Define new scoring criteria based on internal and external data. – Implement the new grid in Python, with parameterizable and reusable logic. – Assist the DevOps team in implementing the real-time grid on IBM ODM, ensuring real-time prediction for customer scoring. – Conduct A/B tests to compare old and new grids on a customer sample. – Key result: 19% performance improvement over the previous version. – Tools: Python, Pandas, Scikit-learn, Pyodbc, Mlflow, IBM ODM. • AI Financial Assistant: – This project aims to develop an AI assistant that helps evaluate companies based on their financial information by calculating financial ratios and providing intelligent insights through natural language interactions with financial documents. – Participate in meetings with client business analysts to better understand their needs. – Develop Python code to extract and clean financial data from documents using OCR. – Build a vector database for efficient storage and retrieval of financial information. – Design an artificial intelligence system capable of interacting with users as an intelligent agent. – Create a user interface allowing end users to interact with the data and receive financial analysis. – Tools: LanGraph, LLM, Python, ChromaDB, Github, MongoDB, OCR.

• AI Sales Assistant (Velbot): – This project consists of creating an AI chatbot using advanced RAG techniques to help users discover the most suitable offers, particularly smartphones currently available on the Orange Senegal website. – Participate in meetings with client business analysts to better understand their needs. – Develop Python code to extract and clean data from websites. – Build a vector database for this cleaned data. – Design an artificial intelligence system capable of interacting with users as an intelligent agent. – Create a user interface allowing end users to interact with the data. – Tools: LangChain, LLM, Python, LanGraph, ChromaDB, Selenium, Github, OpenAI API.

• Customer Appetite for Digital Channel Activation and Usage: – Analysis of customer activation and usage of digital products, development of machine learning algorithms to predict their appetite for using digital channels. – Attend meetings with Business Analysts to understand business needs. – Develop SQL scripts to extract data from sources. – Develop Python code to perform ETL tasks from source to target database. – Clean tables and perform data exploration. – Develop features and target variable. – Design classification algorithms and select optimal classification model. Evaluate models using evaluation metrics (Accuracy, Precision, Recall, Confusion matrix...). – Analyze market test results to validate the developed model. – Tools: Oracle Database, PostgreSQL, PYTHON, SQL, Machine Learning, TensorFlow, Deep Learning, Docker, GitLab. • Customer Attrition Prediction: – This project develops a machine learning algorithm to predict customers at risk of attrition by analyzing their historical data. It includes data exploration, feature engineering, and model development, helping the company retain customers and improve their satisfaction. – Attend meetings with Business Analysts to understand business needs. – Develop SQL scripts to extract data from sources. – Develop Python code to perform ETL tasks from source to target database. – Clean tables and perform data exploration. – Develop features and target variable. – Design classification algorithms and select optimal classification model. Evaluate models using evaluation metrics (Accuracy, Precision, Recall, Confusion matrix...). – Analyze market test results to validate the developed model. – Tools: Oracle Database, PostgreSQL, PYTHON, SQL, Machine Learning, TensorFlow, Deep Learning. • Company Portfolio Valuation: – This project uses NLP techniques to link records between databases without common keys, relying on their textual content. It aims to integrate and analyze data from disparate sources, improving reconciliation processes (Data Matching). – Attend meetings with Business Analysts to understand business needs. – Develop SQL scripts to extract data from sources. – Develop Python code to perform ETL tasks from source to target database. – Develop Python code to perform record matching between two tables. – Tools: Oracle Database, PostgreSQL, PYTHON, SQL, Machine Learning, SQLAlchemy, RecordLinkage, text similarity, text cleaning. • AI Chatbot: – This project develops an AI chatbot using NLP to answer user questions by interacting with documents and websites. It offers a user-friendly interface with features like voice interaction and question history, facilitating accurate and intuitive information retrieval. – Attend meetings with Business Analysts to understand business needs. – Extract data from scanned documents and transform each document to text. – Build a vector database for these text documents. – Use Large Language Models (LLM) to interact with text data extracted from different sources. – Create a user interface inspired by the ChatGPT interface. – Tools: OpenAI API, LangChain, OCR, Large Language Models (LLM), Vector Databases, Embeddings.