consisting of several thousand hours of hand-labelled data, a number of powerful servers (over 70 GPU cards, HPC), we are building large acoustic models, language models, and dictionaries in many languages. Our main goal is to ​​constantly improve the quality of recognition. Using our technologies we have created a number of interesting products, mainly in the SaaS model, for automatic speech recognition (smartphone/desktop apps), process automation, Conversational Intelligence, voicebots, and even offline solutions for smart homes - wherever there is a need for speech recognition and language understanding our technology can be applied and used. VoiceLab solutions are used by thousands of people.