Firas Nafti

• Designed and implemented a LightGBM model for forecasting Estimated Time of Arrival (ETA) with enhanced accuracy, leveraging Databricks for scalable computation. • Refactored existing Gen1 R code reports into optimized Python scripts, improving maintainability and performance in the Gen2 pipeline. • Utilized SQL for efficient data extraction and transformation from diverse sources, ensuring seamless integration into the machine learning pipeline. • Built an end-to-end pipeline encompassing data extraction, cleansing, feature engineering, model training, and deployment, ensuring robust and reproducible workflows. • Leveraged Databricks’ Delta Lake for efficient data management and implemented DBX components to orchestrate and automate workflows. • Tuned the LightGBM model through techniques such as hyperparameter optimization and cross-validation to achieve optimal performance.

• Developed and Implemented Predictive Models: Contributed to creating an XGBoost model to dynamically forecast the (ETA) of shipments in transit and improving prediction accuracy. • Automated Model Training and Deployment: Collaborated with engineering team to achieve weekly automatic model training and deployment. • Statistical Analysis and Modeling: Applied various statistical methods, including bootstrapping, regression analysis, time series analysis, descriptive and inferential statistics; optimized XGBoost models and enhanced performance through A/B testing and experimentation. • Implemented data validation checks to ensure the reliability and quality of input data. • Enhanced Customer Onboarding: Led simulation and reporting development, growing the customer base working with the product team. • Performance Monitoring and Diagnostics: Designed and implemented features like performance monitoring, diagnostic reports, reliability assessments, and a segmentation model to enhance shipment predictions. • End-to-End Automation with CI/CD Pipelines: Contribute to CI/CD pipelines to automate and streamline reliable delivery of features and model updates. • Data Pipeline Optimization: Optimized data pipelines for processing, reducing latency and enhancing predictive model accuracy. • Data Engineering and ETL Process Management: Managed ETL (Extract, Transform, Load) processes to clean, transform, and load large datasets from multiple sources, ensuring high-quality data input for the predictive model. • Release Management: Collaborated with QA on release testing and managed code flow, ensuring seamless feature integration and high code quality. • Stakeholder Communication and Reporting: Developed reports and dashboards to communicate model performance, accuracy, and improvement areas to stakeholders

- Create an inetractive reporting tool to visualize a customer-specific data, this report will be used to extract patterns, trigger issues and will help make conclusions/recommandations. Data used: Customers supply chain data Project: Predictive ETA / Infor Nexus - Science Key words : Supply chain, Time series, R, Python, AWS services