Razvan D.

I build production-grade multi-agent AI systems that actually work, not just prototypes. Under the hood that means hybrid RAG pipelines, graph databases (Neo4j/Memgraph), and the plumbing most demos skip. My focus is on the hard parts: reliable agent orchestration, memory strategies that scale, and deterministic workflows that don't break in production. What I deliver: → Multi-agent systems with proper state management and error handling → RAG and knowledge graphs built for accuracy and low latency → LLM integrations with structured outputs and guardrails → Workflow automation that ties your tools together: REST, OAuth, webhooks, MCP, and third-party apps → Custom MCP servers and Claude builds →†’ Full-stack builds: Python and TypeScript APIs, React and Next.js frontends 8+ years building technology products. Comfortable owning projects end-to-end in async, distributed teams.