Agile delivery of data science projects

Author: Michele Usuelli, Principal Data Scientist, Microsoft

Objectives

Delivering a data science project, some common challenges are:

If you relate to either of these challenges, this article is for you. It will highlight some lessons learned from my projects at Microsoft on how to tackle these problems.

This article shows a delivery methodology which objectives are:

The next sections describe the different components of a proposed solution to these problems.

Personas

Working on a data science project, the starting point is to define the personas involved.

The key personas are the individuals most affected by a data science project:

In the deliver team, the three key technical roles are:

Project set-up

The project step can usual be divided into 3 phases:

The outcome of the project set-up is

Project delivery

After the project is set, the delivery process can be based on the Microsoft TDSP (Team Data Science Process) or on similar methodologies.

TDSP

The TDSP is compatible with the agile delivery methodology.

Per agile project delivery, each piece of work is described by a user stories to be delivered in 2-3 weeks:

Specifically, there are usually three key categories user stories related to data science work

Ultimately, the tool automatically runs the machine learning model and advising the end user about the action to take.

To facilitate the delivery, it is worth defining

Conclusions

This delivery methodology enabled

If you relate with this article, please think about how it can apply to your day-to-day job. If you have any feedback or suggestions, please reach out to me.