Abstract: The successful deployment of machine learning (ML) models into production has traditionally been a complex and resource-intensive process that many organizations struggle with. With the rise of MLOps, a methodology that applies DevOps principles to ML, this process has become much more streamlined. At the Dutch fintech Mollie, we have fully embraced MLOps and implemented a cloud-based ML platform that supports both batch and real-time inference, as well as a suite of MLOps tools to facilitate the entire development cycle.
In this presentation, we will take you on a journey through our MLOps process, from the initial idea to its current use in both development and deployment of the ML models in production. We will demonstrate how our custom platform, built around Google's Vertex AI and other MLOps-related tools (such as Terraform, PyPi, Spark, and Evidently AI), allow us to address the needs of our many stakeholders and streamline the pipeline of ML development, testing, deployment, and monitoring. Moreover, we will share the valuable lessons we learned during the setup of our ML platform, including best practices for model development, version control, and collaboration.
Our presentation will provide practical insights and real-world examples for Data Scientists, ML Engineers, and managers who want to improve their model development process and achieve reliable, scalable, and maintainable deployments. Attendees will leave with a deeper understanding of MLOps and actionable strategies to implement in their own ML projects and organizations.
Bio: Bio Coming Soon!