# How it works
Hydra is designed to be easy, scalable and secure.
# Building models
When you create a model on hydra, you start by selecting a model template. Templates are organized by a data source and a use case type. For example, hydra has templates for
Box form data extraction and
API message classification. Here you have Box or API as the data source and form data extraction (NER) and text classification as the use case types. When you select a template, hydra generates a new model using pre-created code and configuration templates.
During the model generation process, hydra often builds a series of models using various algorithm and optimization templates appropriate for the selected use case type. These models are then scored against each other to find the best performing model. Once the best performing model is found, hydra discards the low performing models and retains the best one.
# Consuming models
Hydra automatically deploys models to a unique and higly-scalable web endpoint. This endpoint is secured using your API key. If you are creating a model using a native application integration such as
Box, this endpoint will be automatically mapped to the source system often using a webhook. So, next time a triggering action takes place in the source system, the hydra model automatically gets invoked and the model writes the predictions to the source (i.e. calling) system. A good example for this form data extraction using Box. Anytime when a form is uploaded to a Box folder, the form data extraction model gets invoked and the model writes the form data fields back to Box as metadata.
If you are using the
API source, hydra provides you a unique model URL, which you can use to make REST API calls for adding training data, training and making predictions.
# Running automations
In most business cases, you want to do something with the predictions — automations can help you with that.
Automations chain a series of actions together to run a number of different tasks automatically after each prediction. Each action step in the automation can use the output of the previous step in addition to the hydra predictions as its input.
Let's take a vendor onboarding automation as an example. A new vendor would upload their onboarding documents through a vendor portal. These documents are automatically placed in a folder in Box (opens new window)— your cloud content management system. You want those documents to be moved to a folder structure created for the new vendor that uses the vendor name as the main folder name. And, you want to notify a series of reviewers that these documents are ready to review.
For this use case, you can use a hydra model for extracting the vendor name from the onboarding documents and then an automation that has two action steps. The first action creates a Box folder layout using the vendor name as the main folder. And, the second action sends out an email to a reviewer email group with a link to the newly created vendor folder. The second action uses the output from the action.