We push the boundaries of AI in natural language processing and computer vision.
What do we do?
We develop AI models in natural language processing and computer vision. Our products and services drive organisations that operate in Poland. We also explore specialised, novel solutions that can be customised for particular industrial sectors or developed specifically for the Polish language.
We are a team of scientists and specialists with experience in scientific research, competitive projects and AI solutions.
- We develop AI methods and models for public and private organisations
- We explore AI methods, primarily in natural language processing and computer vision
- We present our accomplishments at national and international events
- We publish our research findings in scientific journals and as articles on AI
- We share our knowledge on AI
What have we achieved?
Our website showcases our success in neural language models. We are the leaders in the development of language models for the Polish language. Our Polish RoBERTa-large-v2 model outperforms other neural, representative language models, according to the KLEJ ranking.
Based on our own representative models (Polish RoBERTa), we create high-quality sentence-transformer models. we have recently developed top-tier neural models in information retrieval for the Polish language. According to the PIRB benchmark, our advanced text encoders for information retrieval are some of the best on the market.
In December 2023, our general-purpose polish text encoder achieved top place in the MTEB benchmark ranking.
We have also released two generative models for the Polish language: BART and GPT2 with 1.5 billion parameters.
As part of an innovative consortium, we contribute to the development of PLLuM, the first Polish open large language model, which is similar to the GPT models that are fundamental to the success of ChatGPT.
In collaboration with the Information Centre of the Tri-city Academic Computer Network and the Gdańsk University of Technology, we study and develop modern neural language models and their applications. We are in the early stages of pretraining the Polish version of the Llama 2 model.
Knowledge and experience form the cornerstone of our collaboration with PKO BP, the largest Polish bank.
Our projects
the Uniform Antiplagiarism System (JSA)
AI detects plagiarism at higher education institutions in Poland
Collaboration
Do you seek cutting-edge Polish-language processing methods? Do you seek ground-breaking methods for industrial image processing?
We will help you choose and implement AI models and methods that are tailored to your business needs.
Do you want to become our partner?
Send me an email: marek.kozlowski@opi.org.pl
dr inż. Marek Kozłowski
Head of AI LAB
I am the Head of AI LAB at the Polish National Information Processing Institute (OPI). My role involves managing a team of researchers and programmers dedicated to the development of software infused with intelligent methods for processing text and image data. My passions lie in natural language processing, data exploration and machine learning. I am the author of over 45 scientific publications on semantic text processing and machine learning. I have been involved in many commercial machine learning projects. I have collaborated with firms including Samsung, France Telecom, Orange Labs, Millward Brown, Vive Textile Recycling and Connectis. I have taken part in various international machine learning competitions, including the IEEE BigData 2019 Cup.
Follow us on: AI LAB LinkedIn
