AI Analytics Runtime
Artificial Intelligence (AI) pipelines involve multiple specialized roles such as Software Architect, Software Engineer, DevOps Engineer, Data Scientist, IT Systems Administrator, and Big Data Engineer. Data storage and retrieval require a Big Data Engineer, while provisioning and administration of infrastructure fall under the responsibility of an IT Systems Administrator. Data Scientists process and transform data into AI models, and DevOps Engineers and Software Engineers handle deployment and execution of these models in a secure and orchestrated manner. The Software Architect oversees interconnections, technology selection, and architectural decisions. The AI Analytics Runtime component automates the provisioning and deployment of infrastructure and AI models, integrating with other i4 components for functionalities like authentication, monitoring, data acquisition, and marketplace integration.
Benefits
Artificial Intelligence models’ configurations can be specified through a Manifest file depending on the user/model needs: training from historic data, adaptative learning, API configurations, etc
Artificial Intelligence models can be uploaded to be automatically configured, deployed, trained, and run, making them available for generating data predictions
Deployed AI models can be visualized and managed: they can be stopped/started on-demand or by creating a model schedule
New datasets can be uploaded to be used in the training phase of AI model deployments
Dashboard
Visualization of Artificial Intelligence models’ deployments at runtime. Display of some specific model information, ie starting time, the RAM memory in use or its public API URL
Deployer
AI models can be deployed in a Kubernetes cluster by uploading scripts and a Manifest file, creating docker images, training models, generating API endpoints, configuring data transformations, restricting access by user roles, integrating with Wazuh security monitoring, and providing management actions for deployed models.
Scheduler
Deployed AI models can be scheduled to be started or stopped with a specific frequency (ie every Monday) by configuring different options (ie specific time of day), creating an automation of each model run-time
Storage
Dataset files can be uploaded to be used in the training phase of AI models’ deployments. All available datasets are displayed in the upper table. On the other hand, after an AI model is trained, a model file is created and displayed in the lower table. Datasets and trained model files can also be downloaded for other purposes
Additional resources
Training Academy
Software Documentation
Read our easy to follow documentation to learn how to use the i4 Components.