Agentic AI for enterprise systems
AI Engineer / AI Solutions Architect
I design and build AI systems that securely connect LLMs and autonomous agents with enterprise data, ERP systems and real business workflows.
Agentic AI · LLM · MCP · RAG · Enterprise Integration · 1C/ERP
01 / Evidence
Systems explained through constraints, boundaries and engineering decisions—not technology lists.
A production AI agent that gives employees role-aware access to 1C data and business operations through natural language without giving the model unrestricted access to the ERP.
Python · LangGraph · Telegram · MCP
A working R&D framework for specification-driven 1C development with independent AI review, model routing, parallel implementation, verification and human acceptance.
Codex · OpenCode · DeepSeek · 1C
Active R&D on context-aware AI assistants that understand a user's role, work context, enterprise state and desired outcome while operating with bounded autonomy.
Python · LLM · MCP / business tools · 1C
02 / Expertise
AI assistants and agents working with corporate systems through controlled tools, human checkpoints, authorization and audit trails.
Architectures connecting LLM applications with ERP, APIs, corporate data and brownfield systems while keeping data boundaries explicit.
Specification-driven delivery with coding agents, model routing, independent review, testing and evaluation evidence.
03 / Principles
A useful model is only one component. Production systems need explicit responsibility around every tool call and data boundary.
Identity and business permissions must survive the transition from chat interface to enterprise action.
Agents operate through narrow, validated tools instead of unrestricted access to core systems.
Actions, decisions and permission checks need durable evidence that operators can inspect.
Quality is tested against representative tasks and failure modes, not judged from a few convincing demos.
Traces and operational signals make agent behavior diagnosable across models, tools and integrations.
Contracts, timeouts, retries and ownership boundaries matter as much as prompts.
04 / Context
Before focusing primarily on AI systems, I spent more than two decades designing and developing enterprise automation and integration solutions, primarily in the 1C ecosystem. That experience shapes how I treat agents: as participants in corporate information systems, with permissions, contracts, failure modes and responsibility boundaries.