Reducing the Lead Time for Data and AI Projects

Here, we are referring to the time required between the ideation phase of an AI project and the first exposure of an algorithm to a user. We are therefore dealing with a product, not an R&D project.

I insist on this point: an R&D project is not intended to expose a model to a user, but rather trying to solve an open problem. Its deliverable is a scientific paper that explains under which conditions the algorithm does (or does not) answer the question posed (features to be computed, hyperparameters, necessary data, computing capabilities).

An AI product is, first and foremost, software designed to meet a need. The AI algorithm is merely a component, a building block within a broader solution.

In other words, you are not creating a Data product, but software that utilises Artificial Intelligence.

Team Organisation

The smallest unit of work is the team, which must be autonomous and multidisciplinary. It should be predominantly composed of Software Engineers.

The objective? To be able to present the results of a model to multiple users without external assistance.

Our Approach

After a two-week audit to identify the constraints of your product, we commit to delivering a model in production within the first three months. How? By integrating Software Engineers into your team.

A first approach is to temporarily replace your AI module with a dummy algorithm: a programme that takes the same data as input and generates a random result in the expected format.

Deploying it from the start allows for the rapid identification of technical and organisational bottlenecks.

Interested? Contact us!