What if your 99% accuracy model starts performing at 10%? Machine Learning applications are pervasive in modern human activities, but reports show a majority of corporate AI initiatives are struggling to move beyond test stages. Furthermore, when testing is successful, models are typically left running without any control, exposing the system to performance degradation. MLOps (Machine Learning Operations) is a practice for collaboration and communication between data scientists and operations professionals, introducing typical DevOps techniques to this new field. MLOps provides guidelines to the entire development, orchestration, and deployment of machine learning models, allowing to bring effectively models to production and monitor their behavior.
In this webinar Data Reply illustrated which are the challenges of taking a machine learning model in production and how MLOps principles can be applied to solve them in a real business case scenario.
During the webinar, Giuseppe Porcelli, Principal Solutions Architect and Specialist in Machine Learning at AWS, presented how to leverage AWS services to adopt MLOps practices and build a fully-managed ML pipeline.
