datum AI - Home

We are Applied Data Scientists. We are a group of consultants with decades of experience in strategy, management, and IT consulting; as well as software product development, process improvement, organizational change, and technology adoption. We use that baseline to conduct research and development in the data sciences, and we apply our innovative techniques to our client’s needs.

The Pipeline

Machine learning is an iterative process, which should begin with exploration. Do you need it? Should you have it? What's missing? How do you move forward?

Implementing an ML capability is not an event, but a process, an ongoing process at that. Once you decide to move forward there are three basic areas in which you will focus, iteratively over time.

Features | Features are what you feed into machine learning and deep neural network models. They are your data, outside data (e.g., weather, traffic patterns, currency rates, etc.), and a collection of new data designed based on yours and the outside data. Feature engineering, the art and science of creating your features are the first set of steps in the process, and are done in concert with your needs and objectives.
Models | With features in place the process of modeling begins. During this step, features are put through a permutation of differing models in order to create the best learning algorithms. The process may require more features, or eliminate features, but will iteratively work through the modeling and the feature engineering process to get the job done.
Decisions | With the model(s) done the decision support step begins. That is the development of the best way to deliver and consume the models. It could be as simple as running a set of data through the model and reading the output, to creating a more pleasing and sophisticated wrapper around the model and delivering it by way of the cloud. This step may also require organization change and process improvement components.