Samba SG

Mission : Fnac Darty Design and implementation of Fnac Darty’s new marketplace, replacing Mirakl to centralize and automate the management of seller product catalogs. Work Done: • Architecture Workshops and Scoping: Gathering business requirements, defining specifications, and implementing the data architecture. • Data Pipeline Design and Management: Development and orchestration of BigQuery pipelines with Dataform, automated with Airflow, to structure and optimize product catalog management. • Development of Microservices / Jobs: Creation and documentation of FastAPI services in Python for seller product ingestion, validation, and processing, deployed on Cloud Run. • Data Quality Optimization and Integration: Ensuring data consistency and seamless integration with internal tools.

Mission: LVMH Design and implementation of an automated multilingual customer email classification system based on NLP models for Belmond. Work Done: • Data Preparation and Cleaning: Extraction and transformation of customer emails in BigQuery, deduplication, and label standardization. • NLP Model Development: Fine-tuning of Gemini Pro, training of mBERT and XLM-Roberta, and experimentation with classic models XGBoost + TF-IDF for classification in French, English, Spanish, and Portuguese. • Deployment and Automation: Creation of Vertex AI pipelines, implementation of inference APIs on Cloud Functions, and development of a Streamlit interface for batch classification. Result: Improved classification accuracy of customer requests, automation of email processing, and acceleration of the analysis workflow. The final fine-tuned Gemini Pro model achieved an F1-score of 79%.

Mission at L'Oreal : Implementation and enhancement of a comprehensive data catalog (Collibra) to streamline data governance. - Implemented and Maintained complex data management system (DMS) using GCP BigQuery and Collibra. - Implemented microservices using Python (Sanic | Flask) for data ingestion (for different sources) , reading, formatting, and creation. - Implemented datamarts within BigQuery. - Designed diverse dashboards using Looker Studio and Typescript Vue for insights into monitored workflows and metadata. - Created a workflows using Cloud Workflows for process automation and DMS orchestration. - Utilized Cloud Scheduler, App Engine, Cloud Run, Docker, and Cloud Storage for DMS implementation. - Managed a DevOps environment using Terraform, GCP Cloud Build, and GitHub.

Mission at L’Oreal: Implementation of a data consumption layer on GCP to provide easy data access and analysis. Actual work: - Building a data consumption layer on BigQuery by creating a set of tables and views deployed with Terraform. - Setting up data pipelines to ensure seamless integration and deployment with Cloud Build and Workflow. - Maintaining the continuous integration / continuous deployment (CI / CD) process with GitHub. - Contribution to the implementation of row level security (RLS) to protect sensitive data. - Creation of reports in Power BI for stakeholders Result: Improved access and understanding of data through an efficient and reliable data infrastructure in GCP.

Mission : Design and implementation of a real-time multilingual sentiment analysis system for Orange customers on social networks, in particular Twitter. Actual work : - Design and implementation of a Big Data Lambda architecture for data acquisition, extraction and visualisation. - Use of Kafka for real-time data ingestion and Spark/Spark Streaming for continuous data processing. - Development and optimisation of a Machine Learning/Deep Learning pipeline for sentiment analysis in French and Wolof - Using Elasticsearch for storage and Kibana for data visualization. - Deployment on cloud environments (GCP, AWS) for scalability and availability. Result: Contributing to the enhancement of the customer experience by providing real-time sentiment analysis on social networks for a better grasp of customer needs and expectations using Big data technologies #Cloud #GCP #AWS #NLP #BERT

Mission : Development of a web application for the automation of KPI reporting tasks for the project management software JIRA. Actual work : - Data ingestion from the Jira API for a complete view of project progress and performance - Design and development of a MySQL database to store the data - Development of a REST API with SpringBoot to expose the data and make calculations for a front-end use - Development of an interactive front-end with ReactJS for a dashboard to visualise data and KPIs - Containerisation of applications with Docker to facilitate deployments and updates - Application of the Agile Methodology: SCRUM Result: Contributed to productivity improvement by creating an efficient and interactive application to automate KPI reporting tasks for Jira previously based on Excel by leveraging modern technologies.