Core Engine

Website · GitHub · Documentation · Guide

Swiss AI Center contributors

This work is licensed under the AGPL 3.0 license.

Swiss AI Center - 2022-2024 - AGPL 3.0

Introduction

Swiss AI Center - 2022-2024 - AGPL 3.0

Swiss AI Center

  • Five schools from the HES-SO (HEIG-VD, HEIA-FR, HE-Arc, HEVS and HEPIA)
  • Project called Centre Suisse d’Intelligence Artificiel à destination des PMEs (CSIA-PME), also known as the Swiss AI Center.
  • The center’s mission is to accelerate the adoption of artificial intelligence in the digital transition of Swiss SMEs.
Swiss AI Center - 2022-2024 - AGPL 3.0

Features

Swiss AI Center - 2022-2024 - AGPL 3.0
  • Centralize ML services
  • Unify ML services specifications with a HTTP REST API
  • Orchestrate multiple ML services through pipelines
Swiss AI Center - 2022-2024 - AGPL 3.0
  • Beautiful frontend to visualize services and pipelines
  • Extensive documentation available
  • Best practices regarding software development (code reviews, CI/CD)
  • Open source
Swiss AI Center - 2022-2024 - AGPL 3.0

Infrastructure

Swiss AI Center - 2022-2024 - AGPL 3.0

drawing

Swiss AI Center - 2022-2024 - AGPL 3.0

Service specification

Swiss AI Center - 2022-2024 - AGPL 3.0
  • Can be in any language that can implement a REST API
  • Must have the required routes to be “engine” compliant
  • The /compute route must accept the “Task” model
  • Can have its own routes (for specific purposes)
Swiss AI Center - 2022-2024 - AGPL 3.0

Pipeline specification

Swiss AI Center - 2022-2024 - AGPL 3.0

A JSON file containing base information:

  • Name
  • Slug
  • Summary
  • Description
  • Input/Output
  • Tags
Swiss AI Center - 2022-2024 - AGPL 3.0

And a list of “Steps” representing the sequel of services to run with the following data:

  • Identifier (used in the “needs”, “conditions” and “inputs” values)
  • Needs (used to wait until all the services in the array finished their task)
Swiss AI Center - 2022-2024 - AGPL 3.0
  • Condition ([optional] if this specific step should match a condition before being run)
  • Inputs (which data should be put in the entry of the service)
Swiss AI Center - 2022-2024 - AGPL 3.0
Swiss AI Center - 2022-2024 - AGPL 3.0

Next steps

Swiss AI Center - 2022-2024 - AGPL 3.0
  • Pipeline parallelization
  • Toy datasets
  • Functional tests on service declaration
  • And many more…
Swiss AI Center - 2022-2024 - AGPL 3.0

Any questions? 😄

Swiss AI Center - 2022-2024 - AGPL 3.0
Swiss AI Center - 2022-2024 - AGPL 3.0