by Falko Spiller and Stefan Hilmer
Data science systems are based on methods, processes and algorithms for extracting patterns, findings and conclusions from both structured and unstructured data. These systems are very complex to develop and operate. The Stacey matrix makes it clear: complex projects require agile procedures and working methods, which means that the development of data science systems also requires agility.
The desire for agility in system development is very much due to the changing environment. Data science systems and other AI systems must also be able to keep pace with these changes. Corona makes it clear how disruptive change can be. It doesn’t take much imagination to imagine how much the conclusions of a data science system have changed, for example, for the retail sector during the pandemic. But such a system should also always be able to adapt quickly to new circumstances. This can be achieved using the Scrum process framework, for example. In this way, professional agile system development can be achieved.
At the beginning of the development of a data science system, both the problem to be solved and the available data must be considered. The question arises: Can the problem be solved with existing data? This proof can be provided in various ways, whereby pilot developments play a major role. In this phase, an approach based on the design thinking method is recommended.
A distinction can therefore be made between two phases in the development of data science systems. In the first phase, the “project phase”, the team – like pioneers – initially builds the data science system. Only in the second phase, the “product phase”, does the system become part of a professional IT landscape in which it is further developed professionally in an agile environment. In between lies the “point of transition”, the point at which the transition from one phase to another takes place. Timing is particularly important here, as it is important to complete the “project phase” in full and at the same time start the “product phase” at an early stage in order to be able to react to changes quickly enough.
Falko Spiller and Stefan Hilmer will shed light on all of this in their presentation. They will show how agile methods and working methods can be used in the various phases of the development of data science systems in a supportive and targeted manner.
The ARIC workshop is free of charge and will take place online as a webinar on December 7, 2020 at 4 pm. A link to participate will be sent out. Duration: 45 minutes presentation followed by a discussion (approx. 30 minutes).
You can register at the following e-mail address info@aric-hamburg.de.
