Neural networks

Luna Line will reduce the time to create neural networks and the cost of projects by 20-30%.

The MWS AI team has opened access to a large language model for writing computer code. The assistant is based on a language model with 1.5 billion parameters, which can be run even on a laptop to test basic use cases without additional investment.

The Yandex Market team has provided employees at order pickup points with a new service that simplifies the resolution of work issues - "Neuroexpert." It operates in chat mode and helps operators find the necessary information.

K2 NeuroTech (an independent division of K2Tech specializing in supercomputer and AI solutions) presented the PAAC-AI software and hardware complex at the CIPR-2025 conference, which allows for the rapid integration of artificial intelligence into the existing IT environment in accordance with the organization's requirements and security policy. The ready-made solution covers all levels for AI deployment - from computing infrastructure to AI services and user tools.

Artificial intelligence is no longer a toy; it's a tool transforming business reality. How can AI be turned from hype into a tool that saves hundreds of millions of rubles a year? Why does narrow specialization beat OpenAI's billion-dollar budgets? And what should those who are still afraid that algorithms will take away their jobs do? This is an open conversation with Vladimir Chumenko, CEO of aiXpert.

The MWS OctAPI integration platform will allow companies from various sectors of the economy—including industry, healthcare, IT, and others—to implement centralized API management and integrate high-load systems in heterogeneous IT landscapes. According to the developers' estimates, using the platform can reduce the time to launch new products by almost 40%.

Generative artificial intelligence has appeared in the cloud version of Directum Lite, a system for medium and small businesses. The service helps to create the most common documents in the company and save up to 2 hours of working time per day.

Sberbank has introduced an innovative solution for the foreign exchange market - an AI assistant based on generative artificial intelligence. The new system analyzes market data in real-time, identifies anomalies in quotations, and helps traders make more accurate decisions. The implementation of the technology has significantly increased the efficiency of working with foreign exchange transactions and improved the quality of customer service.

Yandex B2B Tech, the corporate division of Yandex, has launched open-source visual-generative models (VLM) on its proprietary ML platform that simultaneously analyze images and text. With VLM models such as Deepseek VL2 Tiny and Gemma3 27B, companies will be able to create product descriptions from photos and quickly find the information they need in documents.

The cold start problem in "machine learning" (ML) is a typical issue faced by systems based on "machine learning." It is particularly relevant for recommendation systems. A neural network trained on a dataset needs to make a prediction about what will best meet the user's needs, but it does not have enough data to do so. Our consultant in preparing this article was Olga Kovaleva, a senior analyst-developer at Yandex.

Test IT, a business unit of "Develonica" (Softline Group), has released a major update to its eponymous test management platform. The key update to TMS in version 5.3 Reticulum is integration with AI models YandexGPT and GigaChat. According to the developers, the AI functionality can reduce the time to generate test scenarios by an average of 30%, and for initial test analysis by 50%.

A new media format has appeared in Avito Advertising for the main page of the branded Rabota service. It is designed to increase brand awareness as an employer.

Telemost has been updated in the Yandex 360 universal set of services for work. Now, corporate clients do not have to spend time compiling "minutes" - summaries of what was agreed upon during a video meeting.

The Yandex Research Artificial Intelligence Laboratory, in collaboration with scientific and technological universities such as HSE University, MIT, KAUST, and ISTA, has developed a method for rapidly compressing large language models (LLMs) without losing quality. The developers emphasized that now, a smartphone or laptop is sufficient to work with the models, eliminating the need for expensive servers and powerful GPUs.

The "MTS Link" platform for business communication, training, and collaboration has received a useful update. In the "Forms" service for tests, surveys, and polls, a neural network has started working, which will automatically check the answers given by respondents in their own words.