Python Snippets to Production-Ready ML Models: Streamline and Trust ML Models in Production – Georgios Gkekas & Dr. Sergei Savin, Vent.io GmbH

Session Outline

With an ever increasing amount of ML models maturing towards production, the question that burdens is; how can we make those models on production more efficient and trustworthy? Long gone are the days of scattered python snippets and messy notebooks. In this talk at the Data Innovation Summit 2024, vent.io presents how they developed a comprehensive ML-Ops strategy to streamline its data science workflows and monitor the ML models. More importantly, Georgios Gkekas and Dr. Sergei Savin show how they leveraged software engineering best practices and a model repository to enhance their code standards, model quality and transparency.

Key Takeaways

• A proper ML-Ops-oriented operating model is key for scaling data science activities
• Treat everything in data science as code and follow software engineering practices
• The importance of a model repository, no matter how trivial it is