Skip to main content

On-Edge & Data Tamed AI

The term 'AI' has come to mean many things; for some, it is used as a catch-all definition for Data Science and Machine Learning. Others use it to refer specifically to the latest evolution of Large Language Models (LLMS) and Generative Pretrained Transformers (GPTs), coupled together as Generative AI.

At DataSapien we utilise a diverse variety of Artificial Intelligence and Machine Learning (AI/ML) tools, including Generative AI, classic machine learning algorithms, and deterministic rules. Each of these varieties of AIML has different strengths that can be utilised in combination to ensure optimal outcomes.

GenerativeAI (or GenAI) transforms the way that we access and process knowledge and data. There is an abundance of AI models for generative text and vision processing, be they Open source or proprietary API based, complex or simple. However, all GenAI models suffer from one thing: hallucinations. Hallucinations are incorrect or misleading results that AI models generate because of insufficient or biased training data. With GenAI, hallucinations are a feature, there will always be a greater than 0% chance of them occurring. DataSapien enables you to sanitise GenAI output based on data it allows you to collect. You can fact-check and validate AI with Zero-Shared Data.

Tiered Intelligence Architecture​

DataSapien employs a three-tiered AI approach.

  1. Deterministic AI that is rule or script-based. You can create data-based Rules and author scripts on Orchestrator and deploy them to be available within minutes to be executed on-edge by DataSapien Mobile SDK.
  2. Classic Probabilistic AI that utilises pretrained AI / ML models. The Orchestrator allows you to configure and deploy models and the DataSapien Mobile SDK contains everything to execute them on-egde, integratable to your use-cases.
  3. Generative AI makes use of on device LLMs (also known as Small Language Models, SLMs or Compact Language Models, CLMs) on-edge with RAG (Retrieval-Augmented Generation) fed from all on-edge collected data held in the DataSapien DataVault.

Thanks to the integrated scripting engine, The DataSapien Mobile SDK provides the functionality to check / filter and sanitize AI output. It is possible to implement freeform accept/ reject scripts or use rules to limit and filter AI generated content.

Scriptable SDK

DataSapien Mobile SDK enables you to run scripts authored on DataSapien Orchestrator. You can employ scripts to interact with AI / ML models deployed from Orchestrator.

Deploy & Execute​

With DataSapien, you have the freedom to pick and deploy any model you see fit. DataSapien Orchestrator allows you to deploy any AI/ML model that can run efficiently on a mobile device. The DataSapien Mobile SDK provides an abstraction layer that enables your app and DataSapien-provided scripts to interact with those models via a generic interface.

With this infrastructure at hand; you can deploy any model and interect with those. You can use models in;

  • Steps in a Journey or Exchange
  • Scripts to collect MeData with special processing
  • Scripts triggered with a data based Rule evaluation
  • Directly from your mobile app