At Fenix Alliance we are helping our customers to build for their users a wide range of solutions that learn and form conclusions with imperfect data.
For over 10 years we've been empowering organizations throgh solutions that interpret the meaning of data, including text, voice, and images.
We seek to unchain everyone's capabilities through solutions that interact with people in natural ways, delivering power across channels.
In order to understand the value that Fenix Alliance's AI & Cognitive Services Practice provides its important to at least have a high level understanding of how Artificial Intelligence models are created.
The following steps summarizes the high-level phases that lead to the creation of useful predictive AI model. For a detailed look at the model building process, please Contact our Technical Support Team.
In this phase, data is collected from sources and prepared for use in training the model. During this phase, the data may be cleaned and de-duplicated, the contents of the data is understood and the data that is the most informative in predicting the outcome is selected.This work is often referred to as data wrangling. Typically the data wrangling is performed by either the data developer or data scientist who have wrote the programs that collect and prepare the data.
Once our data looks fine, the Build Model phase begins, where a subset of the prepared data (which contains both the input and the outcome) is fed into a machine learning or deep learning algorithm to train the model. Then performance of the model is measured against another subset of the prepared data (referred to as the test or evaluation data set), and the model is evaluated on how well it performed in predicting the outcomes described in the test data set.
Assuming the model performed adequately, then the model is saved to a file ready for deployment. In the deploy model phase, the created model file is typically copied to a location where it can be used by the AI application for making predictions. This step is typically performed by one or more of our developers, or more specifically, DevOps engineers who are responsible for making sure the model is always deployed correctly into production and ready to be consumed.