Business Intelligence in a Box

  • Project: BI in a Box
  • Client: INDUST Systems Ltd., Greece
  • Website: bi-in-a-box.gr
  • Description: BI-in-a-Box is an end-to-end Business Intelligence and analytics platform designed to transform raw data into actionable insights. It integrates data from multiple sources (e.g., ERP systems, databases) and provides advanced analytics, reporting, and forecasting capabilities. The platform combines data engineering, statistical modeling, and machine learning techniques to support data-driven decision making. It includes automated data pipelines, interactive dashboards, and time-series forecasting modules, enabling organizations to monitor performance, identify trends, and optimize business operations. Designed with scalability and ease of deployment in mind, the system leverages modern technologies and containerization to deliver a flexible and cost-effective BI solution suitable for organizations of various sizes
  • Contributed to the design and development of BI-in-a-Box, an end-to-end Business Intelligence platform with a focus on time-series forecasting and data-driven decision support. The system implements a scalable ML pipeline for data processing, model training, and inference.
  • Contributions:
    1. Designed and implemented a modular architecture by separating the system into Trainer and Predictor components, enabling scalability, maintainability, and clear MLOps workflows
    2. Developed time-series forecasting pipelines using the Nixtla ecosystem (StatsForecast, NeuralForecast, MLForecast), leveraging both statistical and deep learning models
    3. Built an intelligent model selection framework that evaluates multiple forecasting models using performance metrics (e.g., cross-validation, error metrics) and automatically selects the optimal model
    4. Containerized the application using Docker, ensuring reproducibility, portability, and streamlined deployment
    5. Performed data preprocessing, transformation, and analysis using pandas, and implemented ML workflows with scikit-learn
    6. Optimized training and inference performance, improving model efficiency and pipeline execution

  • Technologies:
    • Python 3.14, Nixtla (StatsForecast, NeuralForecast, MLForecast), Docker, Pandas, Scikit-Learn, Bash shell