Standard Work-Breakdown Structure for AI/ML - tidbit

There are common activities that take place during AI/ML development before it undergoes the process of turning it into a product and deploying it.  Those generic activities could be considered as elements of a standardized WBS for AI/ML development.  They are:

• Task 1 – Understanding the Problem

‒        Problem Definition

• Task 2 – Understanding the Data

‒        Data Acquisition

‒        Data Processing

‒        Data Cleaning

‒        Data Exploration

• Task 3 – Model Development (linear or non-linear)

‒        Model Triage

‒        Feature Vector Engineering

‒        Model Selection

‒        Model Training and Testing

• Task 4 – Model Deployment

‒        Develop Documentation

‒        Define Development stories and Narrative

‒        Development Handoff

‒        Support Development

‒        Support Deployment

‒        In Situ Evaluation and Iteration

‒        Long Term Support

Not all of these tasks and subtasks occur all the time in each and every development stream, but their commonality makes it possible to use them as templates for project management and tracking.  

These tasks and subtasks could be imported into such project management tools as Microsoft Project to reduce the cost of project initiation and setup.

In an analogous manner, these tasks and subtasks could be imported and relabeled as Epic/Feature/User Story in such tools as Jira, Jira Align, Azure DevOps to standardize the development process for AI/ML projects.

Furthermore, taking a baby-step towards the so-called "Trustworthy AI", one could use such things as GUIDs to tie a project to its WBS elements and then to tie the WBS elements to the development artifacts for each task & subtask and package and submit them to Blockchain.  This creates reliable curation of the AI/ML development artifacts and deliverables possible.