Roadmap

When is my feature going to be released?!

The roadmap is an encapsulation of the features that we intend to build for the Core Engine. However, please note that we limited resources and therefore no means of guaranteeing that this roadmap will be followed precisely as described on this page. Rest assured we are working to follow this diligently - please keep us in check!

We intend for this to be a reflection of the current thinking of what we and our users think are the most critical features to be prioritized and added to the Core Engine. If you do not agree with the prioritization, or see a feature here that you expect, please contact us at support@maiot.io, or join the chat in our Discord server.

Feature Roadmap (Updated Sep 1st 2020)

In Progress:

  • Ability to specify custom requirements:

    • Status: Not Started

    • Expected: Q1 2021

  • Ability to add non-tensorflow code in transform functions

    • Status: Not Started

    • Expected: Q2 2021

  • Continuously training pipelines

    • Status: Not Started

    • Expected: Q1 2021 [in active development but completion criteria not defined as of yet]

  • Hyper-parameter Tuning

    • Status: Not Started

    • Expected: Q4 2020

  • Ability to add custom data ingestion component

    • Status: Not Started

    • Expected: Q4 2020

  • Distributed Training

    • Status: Not Started

    • Expected: Q5 2021

  • Ability to add custom split logic

    • Status: Not Started

    • Expected: Q1 2021

Done:

  • Ability to deploy on customers own GCP:

    • Status: Done

  • Images as a datasource:

    • Status: Done

  • Web Interface

    • Status: Done

  • Ability to run custom preprocessing code

    • Status: Done

  • Ability to add a custom model

    • Status: Done

    • Expected: Q2-Q3 2020

  • Ability to run a batch inference pipeline

    • Status: Done

    • Expected: Q2-Q3 2020

History

  • Q4 2020:

    • Transitioned from being a fully managed service to bring your own cloud!

    • Added google orchestration support. No Kubeflow required.

  • Q3 2020:

    • Release of Python SDK

    • Reworked datasource/datasource commits relationship

    • Ability to run different types of pipelines: batch, eval, test and train.

    • Ability to add custom code for models and preprocessing

    • Web interface up and running

  • Q1/Q2 2020:

    • Initial release of Core Engine. CLI interface launched.

    • Support for BigQuery as a datasource

    • Distributed preprocessing pipelines

    • Evaluation of single and multiple pipeline runs

    • Training with GPU

    • Timeseries support