Machine Learning is one of the hottest topics in software engineering today. While it can deliver major competitive advantages, it’s also very challenging: creating the models is just the first step.
One challenge hits particularly hard after your machine learning models have been created; enabling them to work and serve in a production system is a different story. Because ML models don’t age like fine wine, continual updates are always needed. To help you better understand how to design, build, run and manage your ML apps in production, the Lightbend team put together three resources for you - plus a bonus. We hope you find these resources useful! ;)
What it’s about: When it comes to Machine Learning, typically two different groups are responsible for model training and model serving, which introduces inefficiencies. This practical report by Boris Lublinsky, Principal Architect at Lightbend, demonstrates a more standardized approach to model serving and model scoring. By introducing an architecture for serving models in real time as part of input stream processing, data science teams are able to update models without restarting existing applications. [p.s. Boris has also written an additional chapter not included in the eBook on Speculative Model Serving, which you can find here!]
What it’s about: Based on the O’Reilly eBook by Boris Lublinsky mentioned above, this Lightbend webinar digs into different ways to build the right model-scoring solution, using several popular stream processing engines and frameworks such as Python, Beam, Flink, Spark, Kafka streams and Akka.
What it’s about: Most machine learning algorithms are designed to work with stationary data, but in reality, streaming data is rarely stationary. What happens over time is called "Concept Drift", where models based on data collected within a fixed time period begins to suffer from a loss of prediction quality. In this webinar by Emre Velipasaoglu, we look at the benefits of monitoring the model quality by testing the inputs and predictions for changes over time, and using change points in retraining decisions.
As a bonus check out Cloudflow, our tool for rapidly building and running streaming data applications on Kubernetes. Read more about how Cloudflow can help accelerate your journey to the real-time enterprise by simplifying the development, deployment, and operations of complex, multi-component streaming data pipelines. Alternatively, you can request a Cloudflow demo!