We are building a PaaS solution working in a client’s cloud environment.
It’s a smart virtual analyst with NLU interface over BI of any complexity needed for the organization.
“Smart” is the key feature because the analysts’ skills are dynamic and easy-maintainable.
Product uses AI and trains models for phrase detection and slot filling. The solution includes productization of jupyter notebooks as “skills”, usage of business ontologies for data gathering by skills.
We provide generic all-in-one solution for sales, marketing, logistics, etc. Client receives ready-to-use virtual analyst with the chatbot interface integrated to a variety messengers. More skills are implemented by your analytics – smarter is analyst. The skills are easy to come with: starting with few lines of python code you may produce anything that might come handy – graphs, metrics, predictions and insights.
Most of the platform is done in serverless terms. Supported clouds are AWS and Azure, at this time. Most of the components make use of lambdas (function apps), API getaways (API management), sqs (service bus), ecs (aks), DynamoDB (cosmos dB), s3 (blob storage), etc. Dialog capabilities are closely integrated with Amazon Lex service. Jupyter notebooks are productized with docker containers and shipped to execution facility. Infrastructure is able to connect to variety of messengers, e.g., telegram, WhatsApp, teams, email.
Since platform is mostly used for internal company purposes, the accessible data is private and it’s being secured, and additional authorization is applied on messenger layers.