Transformation from strategic use-case conception to use-case industrialization

The second core step of every ML & AI project is the transformation from the strategic conception, developed in the first step, to industrialization. The key challenge in that step is the holistic bundling and evaluation of all potential requirements and use cases that emerge within the organization (as a result of different existing tools and involved departments and users). We have developed a procedure that enables the client to approach the challenge in a structured way and to assess critical indicators for a successful ML project, such as the implementation duration, innovative strength, return on investment and the degree of standardization. This procedure can be extended to include customer-specific criteria and weightings of the individual categories. The outcome is a feature and use case backlog that maps out important milestones of the desired foundation as well as use case specific requirements.

This backlog also provides input for the technical development of the architecture.

Furthermore, we guide the client in analyzing the defined use-cases and identify overlaps. This process allows for a much higher efficiency as the client will be able to reuse some of the mechanisms applied across different use cases. Common overlaps include, for example, similarities in data preparation steps, model types, and requirements for training and optimization, as well as similar deployment procedures in a production environment. The practical use of the models and model-quality monitoring techniques are also often overlapping. Our architects with propose mechanisms of reuse which may include, e.g. the use of a framework of generic classes with project-specific subclasses, or standardization in the form of container-based exchangeable microservices with a uniform interface and a framework that connects the exchangeable components in a uniform way.

In parallel, together with the client, we consider existing tools on the market that would speed up the client’s efforts to achieve the technical goals. The decision of “make vs. buy” is supported by an estimation of the total cost of ownership over a longer period of time (e.g., weighing development costs against license costs).

In summary, this step enables the client to define and build a technical foundation that can efficiently support the complete development and life cycle of ML & AI products