Machine learning is transforming the industry and business, enabling faster and more accurate decision-making. However, maintaining high-performing machine learning models in production environments, often characterized by constantly evolving data, is a significant challenge.

This talk aims to explore the fundamental techniques for monitoring and retraining ML models to ensure consistent performance over time. Through a practical example, we will demonstrate how to apply these solutions to an image-based machine learning model used for fruit defect recognition, highlighting the importance of a structured approach to model maintenance.

We will describe the platform behind the developed solution and the technology stack employed, entirely based on Python libraries, partly open-source and partly developed in-house.

The talk is aimed at developers, data scientists, and ML engineers interested in improving the robustness and reliability of machine learning systems in production.