A meta-learning platform for Kubernetes

"life is a complex nesting of organization at different levels and scales"

NOTE: This project is pre-alpha and not in a working state yet Kubemeta aims to be a compositional way of creating meta-learning environments on Kubernetes that foster AutoML.

apiVersion: kubemeta.ai/v1alpha1
kind: Task
metadata:
  name: dog-classifier
  labels:
    type: image
spec:
  description: classify dogs in images by breed
  input:
    image:
      type: text
      format: binary
  output:
    breedId:
      type: integer

A Task defines a piece of work as an input and output described in OpenAPI format.

A controller will learn to complete Tasks using meta-learning algorithms that create machine learning solutions as Kubernetes objects.

The easiest way to test is using kind. Follow the instructions on the main page to get kind working.

Once kind is installed a cluster can be created using:

make cluster-up