Description
Machine Learning (ML) projects are gaining ground in industry, yet 87% never pass the experiment phase. So often we pour endless time and money building models to help us make predictions about the world, but even top performing models cannot deliver on their intended purpose if they are not able to efficiently run in their intended environment. Machine Learning Operations (MLOps) strives for automation and monitoring at all steps of ML system development and construction so projects have a greater chance of becoming operationalized. With MLOps we have the opportunity to apply traditional DevOps principles that have transformed the world of software delivery, like unification of system development and operations, to machine learning. This talk will cover the principles of MLOps and outline the path to making ML in production a reality.