Skip to content
AI Portfolio

Architect of Autonomous Systems

I design systems where AI agents design, build, evaluate, and ship software autonomously — with built-in checks at every stage.

DevOps → Tech Lead → Architect → CTO → AI. 15 years building systems that replace manual work with automation.

View CasesMy Approach
~$5.5M
economic effect — PGK, Russia's largest private freight operator
55%
overdue → 15% in 6 months — PARMA, 500-person IT holding
500+
staff adopted AI tools, −50% routine work
20
cases with deep dive
Architecture

Software 3.0 — Live Architecture

🎛️ORCHESTRATOR92 services
22 projects · event-driven
🧬RAG + Knowledge Graph
Qdrant · Neo4j · BM25 · Rank Fusion
🗳️2-Stage Deliberation
4 LLMs × 2 rounds · contradiction analysis
📡Monitoring + Grafana
3 dashboards · TG alerts · < 5min detection
🔌92 Integrations
4 MCP servers · Telegram · Gmail · GitHub
Consensus Engine

Deliberation, Not Voting

Every decision passes through 4 independent LLMs. Not voting — contradiction analysis between models.

🧠OpenAI o4-mini
✓ APPROVED
🔮Claude Opus
✓ APPROVED
💎Gemini 2.5 Pro
✗ REJECTED
🌊DeepSeek R1
✓ APPROVED
3/4 ✓Consensus · 2-stage deliberation
Benchmarks

Setup vs Industry Average

Based on Claude Code usage analytics. Industry average estimated from public benchmarks and community data.

Sessions / day34
MY SETUP
INDUSTRY AVG~3
Subagents / day53
MY SETUP
INDUSTRY AVG~1
Tool calls / day1,014
MY SETUP
INDUSTRY AVG~80
Projects shipped22
MY SETUP
INDUSTRY AVG~2–3
Skills20+
MY SETUP
INDUSTRY AVG~2
Hooks per event4
MY SETUP
INDUSTRY AVG~0–1
MCP servers8
MY SETUP
INDUSTRY AVG~0–1
Memory files36
MY SETUP
INDUSTRY AVG~1 (CLAUDE.md)

* My data — Grafana + auto-tracking. Industry — Anthropic Usage Reports, Claude Code community, public benchmarks 2025–2026.

Cases

AI and Enterprise. Two tracks. 20 cases.

Consensus Pattern

Multi-Model Consensus

4 LLM providers × 2-stage deliberation — 28% disagreement rate catches errors single-model misses

3 мес · solo

PythonasyncioOpenAI API
Deep dive →
DAG Orchestration + Kill Gates

Autonomous Code Factory

Multi-agent pipeline with kill gates — Flutter app of 25-40 files in 15-30 min, 3× faster than sequential generation

5 мес · solo

PythonasyncioClaude CLI
Deep dive →
Integration Hub

92 Integrations, One Terminal

MCP infrastructure for AI agents: 11 servers, 6 LLM APIs, 62 Google APIs — agents work with any external service

9 мес · solo

Claude CodeMCP ProtocolPython
Deep dive →
All 20 cases→
Real Metrics

Not a resume. A dashboard.

A Grafana dashboard monitors every session, every tool call, every dollar. Here's today's data.

~80%
Code by agents
Architect & verifier. Same ratio as Karpathy & Cherny.
$0.01
Per deliberation
DeepSeek for volume, premium for critical. 2 stages ~$0.03.
553
Tests, 0 failures
11 services. Bug fix = test first, then fix.
<5min
Detection time
Grafana + TG alerts. Drift visible before complaints.
Grafana · Live Dashboard
Agent Operations Center

Agent Operations Center

Daily KPIs

Daily KPIs

Industry Benchmarks

Industry Benchmarks

Performance & Quality

Performance & Quality

Detailed Percentiles

Detailed Percentiles

Monthly Totals

Monthly Totals

Session Summary

Session Summary

Live Grafana dashboard · Updated every session

Ready to discuss?

If you need an architect who builds autonomous AI systems — reach out.

Serbia-based · CET/CEST timezone · EU-aligned working hours · International contracts experience

Message on TelegramEmailCareer