RMS & RTR — the new architecture of business regulation in Kazakhstan.
The project automates the Risk Management System (RMS) and the Register of Mandatory Requirements (RTR), creating a transparent, data-driven, and predictable model of state control.

By integrating 30 government information systems and 269 risk indicators, inspections are now generated automatically — without human bias.
The Register consolidates all regulatory norms and notifies businesses about changes relevant to their economic activity (OKED).

The result — less administrative burden, more trust, and higher efficiency in the relationship between government and business.

Kazakhstan’s government launched a nationwide Digital Transformation Project to modernize public administration through business process reengineering (BPR) and full-scale digital integration across all sectors.

The project aims to shift from fragmented, paper-based procedures to data-driven, proactive, and citizen-centric governance.
Through reengineering of 600+ government services and the creation of unified digital platforms, the state ensures transparency, efficiency, and interoperability between agencies.

Key components include:

• Process reengineering — simplification and automation of workflows based on user experience;

• Unified data architecture — integration of national registers and analytics systems;

• Platform approach — “Government for Business” and “Government for Citizens” ecosystems;

• AI-based risk management and proactive services — reducing bureaucracy and enabling real-time decision-making.
  
  

The result: faster public services, reduced regulatory burden, and a government that operates as a single digital platform — smart, efficient, and people-oriented.

Key Achievements:

• AI Competence Center: Established the first national Competence Center to develop and implement applied AI solutions for the public and private sectors in Kazakhstan.

• LLM Development & Training: Led the development of KazLLM based on the Qwen2.5–3B architecture, outperforming existing models like BERT by 5-10% in context understanding and text generation.

• Big Data Corpus Construction: Curated a massive high-quality dataset of over 1.3 billion words and 1.1 billion tokens from 40 diverse sources, including government archives, media, and scientific literature.

• Advanced OCR Digitization: Directed the digitization of 28,000+ books from the National Library using the PaliGemma model to convert physical archives into machine-readable data.

• Technical Optimization: Implemented Triton and Flash Attention 2 to reduce GPU memory consumption by 40%, significantly accelerating the model training process.

• Commercialization & Growth: Secured contracts totaling over 294 million KZT (~$560,000) with IT companies and educational platforms, ensuring the project's financial sustainability through 2029.

2. Projects Section

National Large Language Model (KazLLM) World Bank "Productive Sector Consortia I" Grant Program

• Intellectual Property: Filed an international patent application (PCT/KZ2024/000032) for a unique "Method and System for Processing and Generating Machine-Readable Text in Kazakh".

• Scientific Research: Authored research papers submitted to high-impact Q1/Q2 journals, including Annals of Data Science and IEEE Transactions on Pattern Analysis and Machine Intelligence.

• User Interface & Product: Launched a real-time AI assistant integrated into the national messenger Aitu and a dedicated crowdsourcing platform for human-in-the-loop data labeling.

• Strategic Impact: Developed a comprehensive "Industry Development Report" approved by QazInnovations, shaping the national policy for AI integration in Kazakhstan