Complex information systems
We design architectures, develop server and client components, integrate systems and support their evolution. Our experience includes Amirig and TAU Telemetry.
Engineering notebook
From investigating a problem and designing the architecture to developing and maintaining the system.
From investigating a problem and designing the architecture to developing and maintaining the system.
We design architectures, develop server and client components, integrate systems and support their evolution. Our experience includes Amirig and TAU Telemetry.
We extract and compare information from documents and spreadsheets, and develop data search and analysis. In MedStock, we work with public procurement data for the medical market.
We design specialised neural models for understanding queries: identifying entities, conditions, operations and their relationships. We create multilingual training datasets with known correct answers, train custom components and fine-tune adapters for pretrained models. We evaluate quality on separate held-out data, analyse errors and refine the architecture. In 2025–2026, we are applying this approach to semantic routing research; use of the results in MedStock is planned.
In CPO IDE, we are developing large language model (LLM) integration through APIs and adapters for different execution mechanisms. We build context from project materials and knowledge, with access and data-volume limits. We are developing AI-agent task orchestration, resource accounting and execution history. We link results to their original tasks and decisions so they can be reviewed and used in further project development.
We create tools for building applications and managing development. Earlier work includes Miracle and LogicProgram; current work focuses on CPO IDE for projects, knowledge and AI agents.
In current development, we use JavaScript, TypeScript and Python, CSS for interface styling and SQL for working with data. We use Electron for desktop applications, Node.js for server components, React and Vue for web interfaces, and Vite for builds.
The stack varies by project: CPO IDE uses Electron and React; TauTelemetry Platform uses Python, Vue and TypeScript; Amirig DCDM uses Node.js, Express and PostgreSQL; Yuqora uses Electron and AutoHotkey v2; ZEDIS contract data processing uses Python and SQLite.
Current work and plans ↗
02 / Expertise
A useful conversation about a system begins with your processes, data and business constraints.
Identify who will use the system, the problem it should solve and the systems it must work with.
Consider data, equipment and existing software as parts of a single system.
Agree on the scope and an outcome that makes the first stage possible to evaluate.
Start with your challenge