Steve Wilkes, Founder & CTO, Striim Changsha Ma, Solutions Architect, Striim The ultimate goal of many machine learning projects is to continuously utilize a model to make real-time predictions based on current data. Unfortunately, there is often an impedance mis-match here. The model was trained on prepared data, that had been filtered, cleansed and feature extracted, whereas the real-time systems operate on raw data. To solve this issue, the data scientist needs to move upstream and prepare the data earlier, before it is written to disk. During this hands-on session you will learn how to: * filter, enrich and otherwise prepare streaming data * land data continuously, in an appropriate format for training a machine learning model * handle model lifecycles, enabling retraining if the model no longer fits the data * integrate a trained model into the real-time data stream to make continuous predictions * visualize the real-time data and associated predictions, and alert on issues. Attendees of this session will leave having successfully built a full end-to-end operational machine learning application, and understand the principles by which this can be applied to their specific use-cases.