Skip to main content

Rules, ML and AI Models

The DataSapien platform is an enabler of on-edge intelligence use cases. The platform features a three tier intelligence architecture:

  1. Rules - these are first level of intelligence and are static in nature
  2. ML Models - conventional probabilistic models designed to work in a specific field
  3. AI Models - Generative AI Models designed to work in a wide range of fields

The DataSapien Orchestrator is a delivery channel for public and private ML / AI models.

Rules​

Rules are the first intelligence layer built on top of MeData and they are defined and maintained in the Orchestrator and evaluated on-edge by the DataSapien Mobile SDK. The strength of rules lies in their reliability, accuracy and certainty, with zero hallucinations. Fantastic for guiding well defined outcomes.

Rules are constructed with a condition using MeData Definition value assertions and an action to be performed when this condition is met. The Mobile SDK evaluates rules on various occasions and see if their condition is met.

ML Models​

Probabilistic models are trained algorithms that learn patterns from data to make predictions or decisions. They are the "product" of applying an ML algorithm (e.g., a logistic regression model or a trained neural network) to a dataset.

They are usefull in a narrower scope are focused on data-driven learning. The ML models improve with more data and training iterations, outputting probabilities or classifications. They required structured input aligned with training.

The Orchestrator allows provisioning trained ML models so that they can be downloaded and executed by the Mobile SDK.

AI Models​

Generative AI models, such as language, vision and audio models generates new, original data instances resembling the training data, often using probabilistic methods to model data distributions (e.g., GANs where a generator creates data and a discriminator evaluates it).

They excel in creativity and innovation and they handle unstructured data well; produce diverse, novel outputs.

The Orchestrator allows provisioning AI models including LLMs so that they can be downloaded and executed by the Mobile SDK.