What is Konan?
Data scientists are experts when it comes to building state of the art machine learning solutions. Moving those models to production to be used and integrated into bigger systems is a different story. A different skill-set is required to deploy and monitor the models as well as provision and maintain the required infrastructure. This need gave a rise to Machine Learning Operations (MLOps); a set of practices that streamline the process of deploying, monitoring and maintaining machine learning models in production. This is where Konan steps in.
Konan is an MLOps platform that aims to decrease time to deployment as well as provide maintenance and monitoring out-of-the-box, all while being language and platform agnostic. It allows users to have their models up and running in production with the bare minimum requirements. Konan allows the team of data scientists to focus on building great machine learning models and not worry about deploying and managing them.
A model in Konan is your trained model containerized using the requirements specified in the Konan documentation. Models are added inside deployments.
A deployment is the entity you communicate with to interact with your production-ready machine learning model. It provides the necessary REST API endpoints that make the model accessible.
Application Programming Interface (API) is an interface provided by your application to allow other apps/programs to communicate with it. The document that describes this interface is called API Specification or API Documentation.
An API document is used to communicate how to use an API by defining the expected inputs and outputs. In Konan, the API documentation is created for deployments, specifying how to interact with the model.
A container registry a place to store container images. When an image is stored or pushed on a container registry, it can be retrieved or pulled from any machine, provided the correct credentials of the registry is private.