Scalable Optimization of Large Systems with MLOps
Despite the promising outcomes from the proof-of-concept phase, a significant number of machine learning projects terminate prematurely, failing to make the transition to production.
In this talk we show how a three step MLOps strategy helps to determine what needs to be done to lift and shift a prototype to the Microsoft Azure cloud.
We address the challenges of introducing MLOps for controlling and optimizing hundreds of industrial air separation plants using a model-based deep reinforcement learning methodology, emphasising the need for parametrization, scalability, and manageability while also taking architecture considerations into account.
- Basic knowledge of machine learning, data management, and MLOps.
Understanding the decision-making factors for implementing MLOps for large systems in the cloud, including data management, data pipeline, ML pipeline, evaluation, and serving.