_300x.jpg)
COVER FEATURE
RISKS OF ARTIFICIAL INTELLIGENCE
Risks and Pitfalls of AI Agents
Artificial intelligence is often presented as merely a “coding assistant”, but actual experience shows otherwise: it is the AI agent upon which the operation of the system shell depends.
Alexander Ogurechnikov (alexogur04@gmail.com), lead Java developer, B2B Antifraud Platform Team at a large Russian retailer (Moscow).
The AI vs. Cybersecurity Dilemma
Every new AI solution introduced to a corporate network is almost instantly compromised or flagged by existing information security infrastructure. Conversely, newly deployed cybersecurity defenses are immediately targeted and bypassed by rapidly evolving, AI-driven cyberattack tools. Where should businesses invest to minimize threats?
Aleksey Gusev (gusev@ccb.ru), Senior Advisor, Centrocredit Bank, Senior Professor, Peoples’ Friendship University of Russia named after Patrice Lumumba and National Research Nuclear University MEPhI (Moscow).
Cognitive Risks of AI
By delegating thinking to artificial intelligence for the sake of convenience, the fintech industry may lose its key distinctions.
Marianna Danilina (m.danilina@fintechru.org), Head of Strategy, Research and Analytics Directorate, Fintech Association (Moscow).
Semantic Layer and Data Catalog in AI Agent Architecture
A large number of failed pilot projects aimed at implementing AI agents has raised doubts about the efficiency of their integration into corporate business processes. However, there are two tools that help move from “language model hallucinations” to the targeted “precise operation” and “flawless performance”: a data catalog and a semantic layer.
Vitaly Kogtev (vkogtev@mail.ru), Data Science Team Lead, Neuronal Network Technologies (Moscow).
PLATFORMS
From Monolithic Store to Domain-Driven Architecture
When an analytics platform is being scaled, it becomes a bottleneck for the business. A need arises to move from a centralized architecture to a domain-driven one based on LakeHouse, and it should be done in a way to retain the system’s integrity and prevent old issues from carrying over.
Leonid Shumsky (LSHumskiy@inno.tech), Data Platform Division Manager, T1 AI (Moscow).
Microservices have Devoured Monolithic Systems. But what About Data?
A new organizational and architectural model is needed to manage continuously changing data distributed between a multitude of sources.
Vladimir Gromov (gromovv@vtb.ru), Deputy Head, Department of Bank-Wide Systems Technology Development, Daniil Kazakov (dmkazakov@vtb.ru), Head of Data Architecture Directorate, Rostislav Dankov (rdankov@vtb.ru), Project Management Director, Data Management Methodology Group, VTB (Moscow).
PROCESS MANAGEMENT
Business Capabilities Chart
Business process modeling encompasses creating models of both the internal structure of processes and process architecture. For the first task, BPMN 2.0 is the industry standard, but there are no satisfactory notations for process architecture modeling.
Anatoly Belaichuk (bell@b-k.ru), President, ABPMP Russian Chapter; Ilya Tomoradze (ilya@ranepa.ru), Assistant Professor, Russian Presidential Academy of National Economy and Public Administration (Moscow).
DBMS
Safe and Long-Lasting Data Retention
The current rates of data growth are increasingly complicating the task of replacing a database in an operating industrial system, so even at the stage of designing new configurations, it is necessary to take into account not only the current capabilities of the proposed replacement, but also its future development. This applies to all characteristics, including reliability and security. Yandex Database was designed with this in mind from the ground up.
Alexander Zevaykin (azevaykin@yandex-team.ru), Head of String Table Development Group, Konstantin Seleznyov (kseleznyov@yandex-team.ru), Developer, Yandex (Russia).
Secure Databases for Critical Information Infrastructure
For users of critical information infrastructures, it is important who controls the lifecycle of a specific supply, how vulnerabilities are addressed, and what means are used to limit the power of a privileged administrator. If there is no engineering answer to these questions, then the openness of the source code by itself does not guarantee security. How modern databases for critical information infrastructures protect data from both insider and external threats, and how is this implemented in Picodata?
Anna Ermanok (anna@picodata.io), CEO, Picodata (Moscow).
AI Agents in Analytics: From Hype to Efficiency
Forget about vibe coding — let's look without illusions at what layers constitute the foundation of an industrial analytics AI agent and why without metadata and role separation it is doomed to hallucinate.
Nikolay Golov (n.golov@postgrespro.ru), Tengri Data Product Manager, Postgres Professional (Moscow).
OPINION
Culture vs. Technology
Customer service constitutes a fundamental problem in business process design, capable not only of nullifying business digitalization efforts, but also of rendering useless the money invested in growth. In many companies, customer service is often structured in such a way that one can get a service or purchase a product not because of it, but in spite of it. Business just should not be developed by watching it from within.
Taras Soroka (taras.soroka@outlook.com), IT Management and Digital Transformation Consultant (Moscow).
OS ACADEMY
Safe Migration of Scientific Repositories
This paper addresses the challenges of migrating large-scale scientific repositories based on DSpace during platform upgrades to new versions. Traditional manual update methods involve direct intervention in the repository's primary database using destructive operations, which carries significant risks of breaking logical links between objects and retaining traces of exploited critical XXE-class vulnerabilities.
Aleksey Bondyakov (aleksey@jinr.ru), Senior Fellow, Andrey Kondratyev (kondratyev@jinr.ru), Junior Fellow, Joint Institute for Nuclear Research (Dubna).
Digital Twins in Cyberphysical Systems
There is a persistent gap today between the technological capability to create digital twins of business processes and actual management practice. As it turns out, 70% of failures are related not to technical limitations, but rather to management maturity: without restructuring roles, processes, and culture, even the most perfect model will remain just a digital shadow. A three-tier architecture with a closed-loop feedback control system, treating the digital twin as a management pipeline, will help bridge this gap.
Pavel Korotenko (pavelkorotenko@rambler.ru), postgraduate student, Ivan Trifonov (trifonov-iv@ranepa.ru), professor, National Economy Department, EMIT Institute, Russian Presidential Academy of National Economy and Public Administration (Moscow).
