Kubeflow project aims to make it easy for everyone to develop, deploy, and manage composable, multi-cloud, and scalable machine learning on Kubernetes. You can run Kubeflow anywhere where you can deploy Kubernetes.
In the meetup we will simulate a multi cloud and multi team company, building a complex machine learning pipeline. Our goal will be to submit a file to a Kaggle competition, predicting the adoption speed of a given pet using images, metadata, text and additional data sources.
To do this we will leverage different technologies and train multiple machine learning algorithms. We will connect to different data sources, use pre-trained models and train our own models both on our Kubernetes cluster. These components will than be orchestrated to a single pipeline that eventually runs an ensemble method and writes a single prediction file.
The lecture will be in English with an Israeli accent