Technology - Tikos

The TIKOS™ Reasoning Platform technology

Proprietary IP

From the ground up, TIKOS™ is built on original research in AI observability, reasoning and explainability.

Regulations 1st Approach

A technology solution that starts with solving the hard questions of AI compliance in regulated sectors.

Expert-in-the-Loop

TIKOS™ leverages organizational know-how through an ‘expert-in-the-loop’ system design.

Open Architecture

TIKOS™ is agnostic to model class, developer framework, tooling and deployment infrastructure.

Enterprise Scale

Using proprietary distributed processing (ReduceBySQ), TIKOS™ can handle high volumes at minimal latencies.

Flexible Deployment

TIKOS™ can be deployed through SaaS by APIs, SDK and platform access; including private cloud.

TIKOS™ creates two data assets for each model: Cases and Context

‘Cases’ enables observability for every decision output

This includes capturing activation path information from deep-neural networks. Cases are then optimised through information minimisation and serialised for efficient Case indexing, searching, retrieval, matching, and adaptation.

This process delivers log level monitoring and observability for individual decision outputs for any model.

‘Context’ enables explainability for every decision output

‘Context’ extends system capabilities from observation to explainability. Model features are combined with relevant domain information and represented in a knowledge graph, or other datastore.

Matched or adapted Cases relating to individual model output decisions are then explained using the Context.

Innovations

TIKOS™ is built on a family of proprietary formal methodologies, mathematical techniques and algorithms designed to work in concert to deploy the system at scale:

Sequential Collection

Data structure for knowledge serialisation & propagation

Synapses Logger

Process for capturing neural activation path information

ReduceBySQ

Distributed processing for Case search, retrieval and adaptation