CTO, Yoda.Digital · Head of AI Ecosystem, Esempla

Chișinău, Republic of Moldova

I build AI-agent systems, MCP servers, automation infrastructure, and the production machinery that keeps "AI magic" from turning into an incident report with a timestamp.

Senior architect. Hands-on builder. I perform controlled demolitions on elegant nonsense.

I care about systems that survive contact with users, bureaucracy, bad networks, legacy code, public institutions, Monday mornings, and other natural disasters.

Deterministic checks are law. LLM judgment is counsel. Humans are sovereign.

That principle shapes how I design AI infrastructure: agents should be useful, observable, auditable, and powerful without being granted the authority of a caffeinated god process.

I design infrastructure where LLMs are not treated as divine oracles, but as useful, dangerous, and slightly overconfident junior coworkers with tool access.

My current work sits at the intersection of:

• AI-agent systems and MCP ecosystems
• Gov-tech and public-procurement data
• Search, analytics, and SEO automation
• Open-source infrastructure
• Production web platforms
• Developer tooling for people who actually ship

"AI-powered" is not an architecture. It is a warning label unless proven otherwise.

Full GitLab control surface for MCP clients. 86 tools, because apparently "just enough" was unavailable.

Moldova's ANSC procurement appeals, decisions, hearings, and PDFs exposed over MCP.

Moldova MTender public-procurement data over MCP. OCDS 1.1.5, red-flag detection, and enough institutional reality to make optimism file a complaint.

Senior-grade Google Search Console and Indexing API CLI for production SEO workflows.

Google Analytics 4 CLI and MCP server. GA4 automation for humans, agents, and the poor souls still opening dashboards manually.

End-to-end encrypted messaging between Claude Code instances. Curve25519 PFS, Ed25519 TOFU, because even agents deserve operational paranoia.

Full catalogue: opensource.yoda.digital

Commercial AI, web architecture, automation, and production systems for clients across Moldova and abroad.

Building and maintaining real systems since 2015. Production first, presentation later.

Moldova's gov-tech backbone: MTender, e-Apostille, eIntegrity, eGuarantee.

I lead the AI ecosystem layer on top of systems where correctness is not a decorative feature.

Rust · TypeScript · Python · MCP · LLM tooling · PostgreSQL · Docker · Linux · GitLab · WordPress under duress · production incident survival

• Build for production, not demos.
• Prefer boring reliability over theatrical cleverness.
• Keep humans in authority loops.
• Make agents useful before making them autonomous.
• Design control surfaces, not black boxes.
• Logs are evidence. Dashboards are witnesses. Vibes are inadmissible.
• Treat AI as leverage, not absolution.
• Never confuse model confidence with system correctness.

• Production-grade MCP servers
• AI agents with explicit authority boundaries
• Local and cloud LLM tooling
• Gov-tech automation
• Procurement intelligence
• Search and analytics automation
• Open-source systems that can survive real users, which remains rude but unavoidable