The MTS Web Services team has developed its own corporate platform for labeling data used in training artificial intelligence. The solution replaces foreign products LabelStudio Enterprise, Prodigy, Supervisely, labelit, and Audino. According to the company, it covers over 99% of typical tasks for labeling text, audio, images, and video.
The platform supports manual and automated labeling using large language models. Workflows can be configured via a visual builder or a built-in AI agent that understands text instructions, answers questions, and helps generate reports on specialist performance.
A single task can be distributed among multiple annotators and AI models. The system compares the results obtained: it accepts matching answers automatically and sends disputed ones to experts. Ready-made datasets can be exported for training and evaluating AI models.
The solution is deployed on the customer’s infrastructure, including for working with sensitive information and critical information infrastructure facilities. The platform is currently undergoing registration in the registry of domestic software.
According to the company, following the transition to the in-house development in 2025, MWS AI’s labeling costs decreased by 25%, with labor savings exceeding 17,000 person-hours. Over the first ten months of 2026, the figure reached 30,000 hours. Dataset preparation accelerated by 30%, and training new annotators began to take 35% less time.
The platform is used internally at MWS AI and has also been implemented at MTT (Exolve). It is available as a standalone product and as part of the MWS AI Agents Platform. Customers can perform labeling on their own or engage company specialists.