The use of complex computational models makes it possible to understand user statements in natural language, but also to identify the sentiment of the statements, to clean and classify the acquired data and to extract specific objects from the text.

The use of machine learning models makes it possible to predict user behaviour and intentions, search for anomalies or group and cluster data. It also means that the accuracy of the answer to a question posed to a trained system can reach around 85%.

Deep transformer networks with multi-task learning provide fast processing of large amounts of data on computational clusters, including unstructured data. They enable automatic annotation of text sets and building of training corpora.

Machine learning algorithms allow the system to automate the process of learning new content, creating a self-learning database of questions and answers that becomes more precise as the queries increase.

Building computational models using Deep Neural Networks, especially transformer-type networks with multi-task learning, allows automating the process of data acquisition, data annotation and model updating, and thus faster implementation of automation solutions.

Artificial intelligence algorithms using deep neural network learning with reinforcement allow a neural network to learn with a relatively small number of user inputs.

The use of artificial neural networks enabled us to build a conversational engine that underpins the conversational interfaces we create, which customers use to automate selected processes.

The deep learning with transfer used in our recommendation engine makes it possible to match one object to another based on the characteristics that define it - not necessarily expressed explicitly, but derived from content analysis.

Deep neural network learning, including generative adversarial networks and recurrent networks enable automatic analysis and image generation, which can be used for example in creating and optimising advertising campaigns.