In this talk, I will train, deploy, and scale Spark ML and Tensorflow AI Models in a distributed, hybrid-cloud and on-premise production environment.
I will use 100% open source tools including Tensorflow, Spark ML, Jupyter Notebook, Docker, Kubernetes, and NetflixOSS Microservices.
This talk discusses the trade-offs of mutable vs. immutable model deployments, on-the-fly JVM byte-code generation, global request batching, miroservice circuit breakers, and dynamic cluster scaling - all from within a Jupyter notebook.
All code and docker images are 100% open source and available from Github and DockerHub at http://pipeline.io.