I'm an AI automation architect for founders and product & engineering leaders. I design complex agents and sub-agents, orchestrate advanced workflows, and wire Claude Code into production pipelines with n8n — turning ambiguous tasks into reliable, observable automation.
// how it works
A simple, repeatable process for turning a vague “can AI do this?” into a monitored, production-grade workflow.
We pin down the trigger, the data sources, and exactly what a good outcome looks like — no ambiguity.
Claude Code reasoning wired into n8n orchestration — with routing, guardrails, Regex filters, and dedup.
Deployed to production with logging, retries, and a daily digest — observable and safe to re-run.
// expertise
Three layers I design end-to-end — from reasoning hierarchies to the plumbing that keeps them running in production.
Hierarchical agent systems where an orchestrator delegates to specialized sub-agents — each with scoped tools, memory, and guardrails.
Conditional routing, retries, and parallel execution — deterministic control flow around non-deterministic models.
Claude Code reasoning wired directly into n8n nodes, databases, and external APIs.
observable · idempotent · production-ready
Relational core
Battle-tested transactional core for state, audit logs, and relational integrity.
Semantic memory
Embeddings storage that powers retrieval-augmented agents and semantic search.
Typed surface
Typed API surfaces that expose agent capabilities to any client, cleanly.
Event ingress
Event-driven triggers that wake workflows the moment something happens.
Async backbone
Durable async backbone that decouples bursty load from steady processing.
// live workflow
Drop in a task and hit run. A real pipeline executes end-to-end — the agent reasons with Groq, routes the work, and each node lights up on actual progress with real latency and token counts.
Trigger Webhook
inbound event
Reasoning Agent
Groq · routing
Custom Node
n8n · processing
Persist / Dispatch
200 OK
// real cases
Personal automations I built and run in production — proof the stack works, not just slides about it.
JS · 33 cities · Regex · dedup → Sheets → Telegram + Gmail · 09:00
A job-vacancy aggregator for the Katowice agglomeration (Śląskie) — for my wife. Four sources via Apify, a custom JS filter, and a personalized daily digest at 9:00 — instead of hours of manual searching.
0 minutes time spent searching — fully automated
View caseVoice message → Groq STT → Groq LLM → Edge TTS → voice reply
A personal voice bot in Telegram: send a voice message, get a spoken reply. Speech-to-text, an LLM, and text-to-speech — fully card-free (Groq + Edge TTS), running in n8n.
$0 / mo card-free stack — Groq free tier + Edge TTS
View case// case studies
Anonymized client results from production agent systems — the ROI behind the architecture, not just the syntax.
E-commerce · Support & Ops
An n8n + Claude orchestration layer that triages every inbound message, drafts the answer, and routes the rest to the right human — instantly, around the clock.
View caseSales · Lead Generation
A self-running outbound engine that enriches, normalizes, and personalizes prospects at scale — no copy-paste, no manual data entry.
View casen8n · Claude orchestration · production-grade
Metrics based on benchmark pilot deployments, optimized sandbox environments, and operational ROI modeling.
// security & privacy
Enterprise architectures where your data stays yours — encrypted, isolated, and never handed to a model for training.
Your data is never used to train OpenAI or Anthropic models. We run on zero-data-retention API tiers, so prompts and outputs aren't logged, mined, or leaked downstream.
Workflow state, agent memory, and audit logs live in encrypted, isolated Postgres instances — one tenant per database, with row-level security and credentials never shared across clients.
Every n8n webhook and custom node runs in a secure, sandboxed environment — signed payloads, scoped secrets, and zero ambient access to other workflows.
Zero data retention · encrypted at rest & in transit · least-privilege by default
// faq
Common questions about AI automation, agents, and how an engagement works.
Autonomous AI agents and n8n workflows that triage, route, and act — wired into Telegram, Gmail, Google Sheets, REST/GraphQL APIs, and databases. Think reliable automation, not chatbots.
Claude Code for reasoning and orchestration, n8n for the workflow engine, Groq (Llama 3.3 70B) for fast inference, and Apify for data collection — card-free wherever possible.
A focused automation — like the Śląskie job aggregator — ships in days, not months. Larger multi-agent systems are delivered in iterative milestones.
It depends on scope. A focused automation can start from a few hundred dollars; complex multi-agent systems are quoted per project. Share a brief and I'll scope it precisely.
Yes. I take projects from a vague “can AI do this?” to a monitored, production-grade system — and explain every moving part in plain language.
I provide high-end B2B automation services — autonomous AI agents and production-grade n8n workflows, designed, built, and run end to end. I partner with founders and teams who want reliable systems that quietly do the work, not slideware.
Based in Poland · working with clients worldwide.