Flexible technology stack
Dynamic AI software solution is designed to support wide range of operating environments as well as delivered via SaaS.
We build on portable platform of Microsoft .NET Core which is state-of-the-art in programming infrastructure.
To provide our functionality we rely on in-house patented genetic coding technology implementation and also utilize various third-party libraries and tools providing Machine Learning and Natural Language Processing.
Bridging modules are used to call out tools which are implemented in other programming languages and frameworks like Java and Python as well as simplify adding more tools to the system.
Our patented genetic coding technology utilizes data from multiple implementations of the same functionality like POS tagger to produce the most accurate processing algorithms so that we continue adding more information providers to improve the system.
Integration and Orchestration
To receive incoming traffic and submit actions to external systems we use customized integration modules.
Our system is capable of processing any kind of text messages from single sentence to large text.
We currently have implemented integration modules to work with e-mails, chats and social media messaging.
Also, integration modules are able to deliver actions made by human operators to our system to gather feedback and facilitate real-time learning and change of behaviour.
Our system is capable to deliver actions on the same channel message was received (e.g. make a reply or put "Like!") or route it to another channel as well as make a record in the corporate CRM providing orchestration for existing information and communication tools in the company.
Integrations are developed for specific needs of customers to provide interface for internal corporate information systems or specific behaviour on communication channels.
The Dynamic AI system was designed with data security in mind.
As our system classify messages and learn from feedbacks automatically, there is no need for third-party human workers to process the data.
Strong access controls for management web-interface and encryption of sensitive data are available.
Integration with corporate single sign-on authentication systems is also supported.
System accumulates knowledge in the specialized in-house database which can have various back-ends like Redis or PostgreSQL. This feature enables real-time learning functionality based on explicit or implicit feedback.
Using information from the knowledge base the system is able to apply automatic tuning to achieve requested level of accuracy based on confidence score. To accelerate learning of the systems, categorized data-set can be uploaded to the database to produce initial scoring data. Database content can be encrypted, either live data, back-ups or both.