As part of our cooperation with our clients, we create and develop dedicated solutions. We use a wide range of technologies, which we tailor to specific needs. Our solutions are unique because we always start with a thorough analysis of needs and end up with a working, unique solution dedicated to solving a specific problem. Often, brainstorming before the start of a project significantly changes the initial assumptions.

We support Logios in the process of developing and maintaining the entire backend architecture along with applications. We are working on Badacz and Redaktor, which are based on PLI, Logios’ proprietary measure for determining the complexity of texts. The solutions are based on the extraction and analysis of linguistic features specific to the Polish language. The applications are available in both free and commercial versions with the option of on-prem installation.

Our collaboration with ING is a journey through technologies ranging from linguistic features to modern solutions based on transformer architecture. As part of this cooperation, we have developed, produced, and maintain the Prostomat application—a solution for automatic text simplification and text quality measurement. The cooperation has been ongoing since 2018, and we have now implemented the second version of Prostomat. in the production environment. The integration is at the level of ING’s central environment, which means that every employee has access to the solution.

In cooperation with EGO , we design and create solutions based on broadly understood machine learning in the field of natural language processing. Together, we have developed a chatbot- CAM for consultants at the Intergenerational Activity Center (CAM). We are jointly creating a project: From data to decisions: An advanced AI system for instant research analysis.
We are responsible for the entire backend of the solutions, from model training to system architecture and application maintenance. In addition, we supported EGO in creating a course commissioned by the Polish Development Fund on the use of GenAI in public administration, in which we presented the basics of generative language models.

During our collaboration with SWPS, we developed a solution that combines unsupervised data analysis techniques with modern text search and generation techniques (RAG). The solution allows us to understand the problems of people (emigrants) and enables conversation on topics contained in the data with control over the generated response. The tool was developed as a result of research project No. 2022/45/B/HS5/00933, funded by the National Science Centre.

We have developed a solution for the Intergenerational Activity Center that supports people working at chatbot- CAM . It is a private assistant that allows you to immediately answer an employee’s question based on information from public websites and internal CAM documents. In addition, the tool integrates external news services and allows the application to summarize current events for seniors. We are currently in the process of implementing the solution into the CAM architecture.

Playground– is our sandbox, where we present the possibilities offered by various machine learning methods. It is a non-profit project based on publicly available information on the internet (as data feeding the operation). Playground uses a number of methods: crawlers, parsers, classifiers, encoder models, reranking models, clustering methods, generative models, and external (but local) APIs to determine the difficulty of an article. The entire solution is an autonomous, distributed, hybrid architecture, linked with semantic indexing and RAG search technology.