— PORTFOLIO
01 — ABOUT
Background & Approach
I spent 6+ years building systems that do not break — enterprise Java, Spring Boot, and distributed microservices at GlobalLogic. That foundation shapes the multi-agent platforms I build today: durable workflows, explicit approval boundaries, local-first memory, observability, and failure-aware execution. Those are engineering constraints, not optional polish.
At GlobalLogic, I architected a config-driven multi-agent execution harness for an internal LLM platform: LangGraph state graphs across 41 agents, deterministic and LLM-routed execution paths, and validation at startup and in CI for unresolvable skills, invalid routing graphs, and cycles. The validator exposed approval gates that could never fire and agents reserving model budget for calls they never made — turning agent configuration into something testable and operable. I apply the same platform mindset to my self-hosted multi-agent development platform for AI-assisted delivery and local infrastructure operations: bounded specialist agents across multiple LLM providers, durable Git-backed local-first knowledge, explicit operator confirmation, and a Node.js, TypeScript, and React operations surface for live infrastructure, agent activity, scheduled automation, and evidence-backed delivery records.
CURRENTLY SEEKING
I am looking for a product company where AI architecture and developer experience are the product — not a thin wrapper around an API call. I thrive designing multi-agent orchestration, evaluation harnesses, MCP-enabled tools, and reliable AI delivery workflows with observability, cost signals, and fault tolerance built in. Remote only. Contract or full-time.
02 — SKILLS
Technical Stack
Protocol Engineering
MCP (Streamable HTTP, RFC 9728 OAuth discovery), A2A (task lifecycle, SSE streaming, multi-turn input-required flows), BYOK architecture, multi-tenant RAG isolation
Agentic Systems
AI SDLC harnesses · bounded multi-agent orchestration · LangGraph stateful workflows · LiteLLM multi-provider routing · durable Redis event queues · explicit approval gates · session and durable memory · n8n workflow automation
Product & Developer Interfaces
Node.js · JavaScript · TypeScript · React 19 · REST APIs · GitHub webhooks · CI/CD · developer workflow surfaces
LLM Integration
Anthropic Claude · OpenAI API · Google Gemini · Groq · Ollama · OpenRouter · multi-model fallback · token cost optimization · structured output · prompt engineering · evaluation harnesses
RAG & Knowledge
ChromaDB, Pinecone, pgvector, FAISS, Neo4j · GraphRAG · embedding optimization · chunking strategies · contextual retrieval · semantic caching (Redis)
Backend Foundation
Python (FastAPI, Pydantic) · Java 17 (Spring Boot 3, Spring Cloud) · Go · PostgreSQL · Redis · Kafka · event-driven architecture · microservices
Infrastructure
Docker · Kubernetes · Oracle Cloud Infrastructure (ARM VPS) · Proxmox homelab · AWS (EC2, RDS, S3, EKS) · Terraform · Ansible · GitHub Actions CI/CD · nginx · Cloudflare
Observability
Prometheus · Grafana · OpenTelemetry · cAdvisor · audit trails · GDPR compliance (right to erasure, data export, retention automation)
Certifications
Neo4j Certified Professional · 10× Anthropic (Claude Code in Action, Building with the Claude API, Claude in Amazon Bedrock, Claude with Vertex AI, Subagents, Agent Skills, and more)
03 — PROJECTS
Selected AI Engineering Work
04 — CONTACT
Let's Talk
Interested in AI SDLC, agentic platforms, or developer tooling? Drop a message — I read everything.
Based in Ukraine. Available for remote contracts, consulting, and full-time roles. Response time: 24–48h.