Marouan Ezbakhe

Built and integrated a complete predictive system into the SmartFellah platform to support farmers with real-time decision-making: • Built alerts prediction model with TensorFlow and deployed it in the ms-ml microservice using FastAPI. • Developed activities prediction model with XGBoostClassifier inside ms-ml for agricultural task recommendations (irrigation, seeding, harvesting, treatment, sales). • Implemented ms-sensor with Airflow DAGs to generate and publish sensor data to Kafka topics. • Built ms-activity and ms-alert microservices to consume Kafka topics, call ms-ml for predictions, and persist results in Postgres. • Integrated predicted results into the React frontend using WebSocket for real-time updates in the UI. • Containerized services with Docker Compose for local integration and testing. Tech stack: TensorFlow, XGBoost, FastAPI, NestJS, React, WebSocket, Apache Kafka, Apache Airflow, Postgres, Docker, Python.