I'm a Solution Architect at EPAM Systems focused on AI-powered systems, agentic architectures, and distributed platform engineering. I design and ship production-grade AI platforms that bridge the gap between ML research and real-world enterprise deployments.
production_rag → multi-tenancy · hallucination detection · cost budgets · audit
agent_orchestration → LangGraph · Anthropic Agent SDK · A2A protocol
mcp_servers → expose tools + data to agents safely
ai_observability → tracing · evals · monitoring at scale
platform → OpenShift / Kubernetes-native AI deploys
AI / ML — LangChain · LangGraph · PydanticAI · Anthropic SDK · MCP · A2A · Ollama · OpenAI · ChromaDB · pgvector · Hugging Face
Backend — Python · Go · TypeScript · FastAPI · Celery · Temporal
Data & Messaging — PostgreSQL · MongoDB · Redis · Cassandra · MinIO
Infra & Cloud — Docker · Kubernetes · OpenShift · Helm · ArgoCD · Terraform · GCP · AWS · Azure
Frontend — React · Vite · Tailwind
🛡️ SentinelRAG · flagship
Multi-tenant, RBAC-aware, evaluation-driven enterprise RAG platform
Production-shaped RAG with pgvector hybrid retrieval, layered hallucination detection, immutable audit, and per-tenant cost budgets. Multi-cloud Terraform, Helm + ArgoCD, k6 + Chaos Mesh resilience tests, and 37 ADRs explaining every non-obvious decision.
Python FastAPI Temporal PostgreSQL pgvector Redis Helm ArgoCD Terraform Kubernetes
Local-first test failure intelligence for CI pipelines and QA teams
Turns raw automated test results (JUnit, pytest, TestNG, Allure, Cypress, Playwright…) into failure intelligence and release-risk signals — GO / CONDITIONAL_GO / NO_GO. Offline-capable, exposed via API / CLI / UI / MCP (48 tools).
Python TypeScript MCP Ollama Docker Compose Kubernetes
Multi-agent QA orchestration platform — 10 specialized agents, one control plane
Orchestrates specialized QA agents with an evaluation harness, multi-tenant RLS isolation, and enterprise controls. 728 unit tests, mypy --strict clean across 250 files.
Python LangGraph PostgreSQL FastAPI Kubernetes
Simulate, test, and validate AI agent integrations — without calling real LLMs or burning tokens
Drop-in mock for OpenAI / Anthropic / Gemini / MCP endpoints in a single Go binary. MCP-conformance-tested, Helm-deployable, with chaos injection for resilience testing.
Go MCP OpenAI API Anthropic API Helm
More AI & agent projects
⚙️ Auto Tests Gen — Agentic SDLC test engineering: PR/JIRA-triggered workflows with OpenHands + Claude Code headless and a deterministic push gate. Python AI Agents OpenShift
🧩 AgentMemory — Guides & diagrams on memory-augmented agent systems: short-term, long-term, episodic, semantic. Mermaid Documentation
🧪 Python AI Agents Demo — Multi-agent demos with MCP server, FAISS RAG, LangGraph human-in-the-loop, and a FastAPI + React trace UI. Python LangGraph MCP
Distributed systems & platform engineering
| Project | What it is | Stack |
|---|---|---|
| 🛡️ Rate Limiter | Token-bucket & sliding-window via Redis Lua, admin control plane with approvals/rollback, OTel | Python Redis |
| 📊 MetricMesh | Time-series anomaly detection — TimescaleDB, Celery, Prophet / IsolationForest, alert routing | Python TimescaleDB |
| 🧪 Test Execution Orchestrator | Distributed test control plane — elastic workers, OTel ingest, statistical flaky detection | Go OpenTelemetry |
| 📅 Distributed Job Scheduler | Two-layer Cassandra + Redis scheduler with visibility timeouts and atomic Lua dequeue | Python Cassandra Redis |
| 💳 Payment System | Transactional flows with Debezium CDC → Kafka, idempotency, AI-assisted ops triage | TypeScript Kafka |
| 🌐 Web Crawler | Distributed crawler — BullMQ queues, robots.txt politeness, management UI | TypeScript |
Professional-level, verified on Credly:
| Certification | Provider |
|---|---|
| Professional Cloud Architect | Google Cloud |
| Professional ML Engineer | Google Cloud |
| Professional Data Engineer | Google Cloud |
| Professional Cloud DevOps Engineer | Google Cloud |
| Solutions Architect – Associate | AWS |
| ML Engineer – Associate | AWS |
Full list of certifications and skill badges → Credly profile
|
|
|
Always open to discussing AI system architecture, agentic workflows, and platform engineering.
"The best architecture is the one that disappears — letting AI do the work."